mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-11 20:09:23 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #25710 from gursimarsingh:improved_object_detection_sample

Browse Source 
Merged yolo_detector and object detection sample #25710

Relates to #25006

This pull request merges the yolo_detector.cpp sample with the object_detector.cpp sample. It also beautifies the bounding box display on the output images

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
This commit is contained in:
Gursimar Singh 2024-09-18 18:49:46 +05:30 committed by GitHub
parent 46b800f506
commit e823493af1
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
6 changed files with 684 additions and 790 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

64
doc/tutorials/dnn/dnn_yolo/dnn_yolo.markdown
View File

@ -150,13 +150,12 @@ Once we have our ONNX graph of the model, we just simply can run with OpenCV's s
3. Run the following command:
@code{.cpp}
./bin/example_dnn_yolo_detector --input=<path_to_your_input_file> \
--classes=<path_to_class_names_file> \
./bin/example_dnn_object_detection <model_name> --input=<path_to_your_input_file> \
--labels=<path_to_class_names_file> \
--thr=<confidence_threshold> \
--nms=<non_maximum_suppression_threshold> \
--mean=<mean_normalization_value> \
--scale=<scale_factor> \
--yolo=<yolo_model_version> \
--padvalue=<padding_value> \
--paddingmode=<padding_mode> \
--backend=<computation_backend> \
@ -166,7 +165,7 @@ Once we have our ONNX graph of the model, we just simply can run with OpenCV's s
@endcode
- --input: File path to your input image or video. If omitted, it will capture frames from a camera.
- --classes: File path to a text file containing class names for object detection.
- --labels: File path to a text file containing class names for object detection.
- --thr: Confidence threshold for detection (e.g., 0.5).
- --nms: Non-maximum suppression threshold (e.g., 0.4).
- --mean: Mean normalization value (e.g., 0.0 for no mean normalization).
@ -191,43 +190,28 @@ To demonstrate how to run OpenCV YOLO samples without your own pretrained model,
Run the YOLOX detector(with default values):
@code{.sh}
git clone https://github.com/opencv/opencv_extra.git
cd opencv_extra/testdata/dnn
python download_models.py yolox_s_inf_decoder
cd ..
export OPENCV_TEST_DATA_PATH=$(pwd)
cd opencv/samples/dnn
export OPENCV_DOWNLOAD_CACHE_DIR=<path to download the model>
cd ../data
export OPENCV_SAMPLES_DATA_PATH=$(pwd)
python download_models.py yolov8x --save_dir=$OPENCV_DOWNLOAD_CACHE_DIR
cd <build directory of OpenCV>
./bin/example_dnn_yolo_detector
./bin/example_dnn_object_detection yolov8x
@endcode
This will execute the YOLOX detector with your camera.
For YOLOv8 (for instance), follow these additional steps:
@code{.sh}
cd opencv_extra/testdata/dnn
python download_models.py yolov8
cd ..
export OPENCV_TEST_DATA_PATH=$(pwd)
cd opencv/samples/dnn
export OPENCV_DOWNLOAD_CACHE_DIR=<path to download the model>
cd ../data
export OPENCV_SAMPLES_DATA_PATH=$(pwd)
python download_models.py yolov8n --save_dir=$OPENCV_DOWNLOAD_CACHE_DIR
cd <build directory of OpenCV>
./bin/example_dnn_yolo_detector --model=onnx/models/yolov8n.onnx --yolo=yolov8 --mean=0.0 --scale=0.003921568627 --paddingmode=2 --padvalue=144.0 --thr=0.5 --nms=0.4 --rgb=0
./bin/example_dnn_object_detection yolov8n --model=onnx/models/yolov8n.onnx --mean=0.0 --scale=0.003921568627 --paddingmode=2 --padvalue=144.0 --thr=0.5 --nms=0.4 --rgb=0
@endcode
For YOLOv10, follow these steps:
@code{.sh}
cd opencv_extra/testdata/dnn
python download_models.py yolov10
cd ..
export OPENCV_TEST_DATA_PATH=$(pwd)
cd <build directory of OpenCV>
./bin/example_dnn_yolo_detector --model=onnx/models/yolov10s.onnx --yolo=yolov10 --width=640 --height=480 --scale=0.003921568627 --padvalue=114
@endcode
This will run `YOLOv10` detector on first camera found on your system. If you want to run it on a image/video file, you can use `--input` option to specify the path to the file.
VIDEO DEMO:
@youtube{NHtRlndE2cg}
@ -238,30 +222,30 @@ module this is also quite easy to achieve. Below we will outline the sample impl
- Import required libraries
@snippet samples/dnn/yolo_detector.cpp includes
@snippet samples/dnn/object_detection.cpp includes
- Read ONNX graph and create neural network model:
@snippet samples/dnn/yolo_detector.cpp read_net
@snippet samples/dnn/object_detection.cpp read_net
- Read image and pre-process it:
@snippet samples/dnn/yolo_detector.cpp preprocess_params
@snippet samples/dnn/yolo_detector.cpp preprocess_call
@snippet samples/dnn/yolo_detector.cpp preprocess_call_func
@snippet samples/dnn/object_detection.cpp preprocess_params
@snippet samples/dnn/object_detection.cpp preprocess_call
@snippet samples/dnn/object_detection.cpp preprocess_call_func
- Inference:
@snippet samples/dnn/yolo_detector.cpp forward_buffers
@snippet samples/dnn/yolo_detector.cpp forward
@snippet samples/dnn/object_detection.cpp forward_buffers
@snippet samples/dnn/object_detection.cpp forward
- Post-Processing
All post-processing steps are implemented in function `yoloPostProcess`. Please pay attention,
that NMS step is not included into onnx graph. Sample uses OpenCV function for it.
@snippet samples/dnn/yolo_detector.cpp postprocess
@snippet samples/dnn/object_detection.cpp postprocess
- Draw predicted boxes
@snippet samples/dnn/yolo_detector.cpp draw_boxes
@snippet samples/dnn/object_detection.cpp draw_boxes

27
samples/dnn/common.hpp
View File

@ -12,6 +12,8 @@ std::string findFile(const std::string& filename);
std::string findModel(const std::string& filename, const std::string& sha1);
std::vector<std::string> findAliases(std::string& zooFile, const std::string& sampleType);
inline int getBackendID(const String& backend) {
std::map<String, int> backendIDs = {
{"default", cv::dnn::DNN_BACKEND_DEFAULT},
@ -177,8 +179,33 @@ std::string genPreprocArguments(const std::string& modelName, const std::string&
modelName, zooFile)+
genArgument(prefix + "labels", "Path to a text file with names of classes to label detected objects.",
modelName, zooFile)+
genArgument(prefix + "postprocessing", "Indicate the postprocessing type of model i.e. yolov8, yolonas, etc.",
modelName, zooFile)+
genArgument(prefix + "sha1", "Optional path to hashsum of downloaded model to be loaded from models.yml",
modelName, zooFile)+
genArgument(prefix + "download_sha", "Optional path to hashsum of downloaded model to be loaded from models.yml",
modelName, zooFile);
}
std::vector<std::string> findAliases(std::string& zooFile, const std::string& sampleType) {
std::vector<std::string> aliases;
zooFile = findFile(zooFile);
cv::FileStorage fs(zooFile, cv::FileStorage::READ);
cv::FileNode root = fs.root();
for (const auto& node : root) {
std::string alias = node.name();
cv::FileNode sampleNode = node["sample"];
if (!sampleNode.empty() && sampleNode.isString()) {
std::string sampleValue = (std::string)sampleNode;
if (sampleValue == sampleType) {
aliases.push_back(alias);
}
}
}
return aliases;
}

56
samples/dnn/models.yml
View File

@ -16,9 +16,9 @@ yolov8x:
width: 640
height: 640
rgb: true
classes: "object_detection_classes_yolo.txt"
background_label_id: 0
sample: "yolo_detector"
labels: "object_detection_classes_yolo.txt"
postprocessing: "yolov8"
sample: "object_detection"
yolov8s:
load_info:
@ -30,11 +30,11 @@ yolov8s:
width: 640
height: 640
rgb: true
classes: "object_detection_classes_yolo.txt"
background_label_id: 0
sample: "yolo_detector"
labels: "object_detection_classes_yolo.txt"
postprocessing: "yolov8"
sample: "object_detection"
yolov8n:
yolov8:
load_info:
url: "https://github.com/CVHub520/X-AnyLabeling/releases/download/v0.1.0/yolov8n.onnx"
sha1: "68f864475d06e2ec4037181052739f268eeac38d"
@ -44,9 +44,9 @@ yolov8n:
width: 640
height: 640
rgb: true
classes: "object_detection_classes_yolo.txt"
background_label_id: 0
sample: "yolo_detector"
labels: "object_detection_classes_yolo.txt"
postprocessing: "yolov8"
sample: "object_detection"
yolov8m:
load_info:
@ -58,9 +58,9 @@ yolov8m:
width: 640
height: 640
rgb: true
classes: "object_detection_classes_yolo.txt"
background_label_id: 0
sample: "yolo_detector"
labels: "object_detection_classes_yolo.txt"
postprocessing: "yolov8"
sample: "object_detection"
yolov8l:
load_info:
@ -72,8 +72,8 @@ yolov8l:
width: 640
height: 640
rgb: true
classes: "object_detection_classes_yolo.txt"
background_label_id: 0
labels: "object_detection_classes_yolo.txt"
postprocessing: "yolov8"
sample: "yolo_detector"
# YOLO4 object detection family from Darknet (https://github.com/AlexeyAB/darknet)
@ -90,7 +90,7 @@ yolov4:
width: 416
height: 416
rgb: true
classes: "object_detection_classes_yolo.txt"
labels: "object_detection_classes_yolo.txt"
background_label_id: 0
sample: "object_detection"
@ -105,7 +105,7 @@ yolov4-tiny:
width: 416
height: 416
rgb: true
classes: "object_detection_classes_yolo.txt"
labels: "object_detection_classes_yolo.txt"
background_label_id: 0
sample: "object_detection"
@ -120,7 +120,7 @@ yolov3:
width: 416
height: 416
rgb: true
classes: "object_detection_classes_yolo.txt"
labels: "object_detection_classes_yolo.txt"
background_label_id: 0
sample: "object_detection"
@ -135,24 +135,10 @@ tiny-yolo-voc:
width: 416
height: 416
rgb: true
classes: "object_detection_classes_pascal_voc.txt"
labels: "object_detection_classes_pascal_voc.txt"
background_label_id: 0
sample: "object_detection"
yolov8:
load_info:
url: "https://github.com/CVHub520/X-AnyLabeling/releases/download/v0.1.0/yolov8n.onnx"
sha1: "68f864475d06e2ec4037181052739f268eeac38d"
model: "yolov8n.onnx"
mean: [0, 0, 0]
scale: 0.00392
width: 640
height: 640
rgb: true
postprocessing: "yolov8"
classes: "object_detection_classes_yolo.txt"
sample: "object_detection"
# Caffe implementation of SSD model from https://github.com/chuanqi305/MobileNet-SSD
ssd_caffe:
load_info:
@ -165,7 +151,7 @@ ssd_caffe:
width: 300
height: 300
rgb: false
classes: "object_detection_classes_pascal_voc.txt"
labels: "object_detection_classes_pascal_voc.txt"
sample: "object_detection"
# TensorFlow implementation of SSD model from https://github.com/tensorflow/models/tree/master/research/object_detection
@ -183,7 +169,7 @@ ssd_tf:
width: 300
height: 300
rgb: true
classes: "object_detection_classes_coco.txt"
labels: "object_detection_classes_coco.txt"
sample: "object_detection"
# TensorFlow implementation of Faster-RCNN model from https://github.com/tensorflow/models/tree/master/research/object_detection

714
samples/dnn/object_detection.cpp
View File

@ -1,68 +1,114 @@
//![includes]
#include <fstream>
#include <sstream>
#include <opencv2/dnn.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/imgcodecs.hpp>
#include <opencv2/highgui.hpp>
#if defined(HAVE_THREADS)
#define USE_THREADS 1
#endif
#ifdef USE_THREADS
#include <mutex>
#include <thread>
#include <queue>
#endif
#include "iostream"
#include "common.hpp"
std::string param_keys =
"{ help h | | Print help message. }"
"{ @alias | | An alias name of model to extract preprocessing parameters from models.yml file. }"
"{ zoo | models.yml | An optional path to file with preprocessing parameters }"
"{ device | 0 | camera device number. }"
"{ input i | | Path to input image or video file. Skip this argument to capture frames from a camera. }"
"{ framework f | | Optional name of an origin framework of the model. Detect it automatically if it does not set. }"
"{ classes | | Optional path to a text file with names of classes to label detected objects. }"
"{ thr | .5 | Confidence threshold. }"
"{ nms | .4 | Non-maximum suppression threshold. }"
"{ async | 0 | Number of asynchronous forwards at the same time. "
"Choose 0 for synchronous mode }";
std::string backend_keys = cv::format(
"{ backend | 0 | Choose one of computation backends: "
"%d: automatically (by default), "
"%d: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit), "
"%d: OpenCV implementation, "
"%d: VKCOM, "
"%d: CUDA }", cv::dnn::DNN_BACKEND_DEFAULT, cv::dnn::DNN_BACKEND_INFERENCE_ENGINE, cv::dnn::DNN_BACKEND_OPENCV, cv::dnn::DNN_BACKEND_VKCOM, cv::dnn::DNN_BACKEND_CUDA);
std::string target_keys = cv::format(
"{ target | 0 | Choose one of target computation devices: "
"%d: CPU target (by default), "
"%d: OpenCL, "
"%d: OpenCL fp16 (half-float precision), "
"%d: VPU, "
"%d: Vulkan, "
"%d: CUDA, "
"%d: CUDA fp16 (half-float preprocess) }", cv::dnn::DNN_TARGET_CPU, cv::dnn::DNN_TARGET_OPENCL, cv::dnn::DNN_TARGET_OPENCL_FP16, cv::dnn::DNN_TARGET_MYRIAD, cv::dnn::DNN_TARGET_VULKAN, cv::dnn::DNN_TARGET_CUDA, cv::dnn::DNN_TARGET_CUDA_FP16);
std::string keys = param_keys + backend_keys + target_keys;
//![includes]
using namespace cv;
using namespace dnn;
using namespace std;
float confThreshold, nmsThreshold;
std::vector<std::string> classes;
const string about =
"Firstly, download required models using `download_models.py` (if not already done). Set environment variable OPENCV_DOWNLOAD_CACHE_DIR to specify where models should be downloaded. Also, point OPENCV_SAMPLES_DATA_PATH to opencv/samples/data.\n"
"To run:\n"
"\t ./example_dnn_object_detection model_name --input=path/to/your/input/image/or/video (don't give --input flag if want to use device camera)\n"
"Sample command:\n"
"\t ./example_dnn_object_detection yolov8 --input=$OPENCV_SAMPLES_DATA_PATH/baboon.jpg\n"
inline void preprocess(const Mat& frame, Net& net, Size inpSize, float scale,
const Scalar& mean, bool swapRB);
"Model path can also be specified using --model argument. ";
void postprocess(Mat& frame, const std::vector<Mat>& out, Net& net, int backend);
const string param_keys =
"{ help h | | Print help message. }"
"{ @alias | | An alias name of model to extract preprocessing parameters from models.yml file. }"
"{ zoo | ../dnn/models.yml | An optional path to file with preprocessing parameters }"
"{ device | 0 | camera device number. }"
"{ input i | | Path to input image or video file. Skip this argument to capture frames from a camera. }"
"{ thr | .5 | Confidence threshold. }"
"{ nms | .4 | Non-maximum suppression threshold. }"
"{ async | 0 | Number of asynchronous forwards at the same time. "
"Choose 0 for synchronous mode }"
"{ padvalue | 114.0 | padding value. }"
"{ paddingmode | 2 | Choose one of padding modes: "
"0: resize to required input size without extra processing, "
"1: Image will be cropped after resize, "
"2: Resize image to the desired size while preserving the aspect ratio of original image }";
void drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame);
const string backend_keys = format(
"{ backend | default | Choose one of computation backends: "
"default: automatically (by default), "
"openvino: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit), "
"opencv: OpenCV implementation, "
"vkcom: VKCOM, "
"cuda: CUDA, "
"webnn: WebNN }");
void callback(int pos, void* userdata);
const string target_keys = format(
"{ target | cpu | Choose one of target computation devices: "
"cpu: CPU target (by default), "
"opencl: OpenCL, "
"opencl_fp16: OpenCL fp16 (half-float precision), "
"vpu: VPU, "
"vulkan: Vulkan, "
"cuda: CUDA, "
"cuda_fp16: CUDA fp16 (half-float preprocess) }");
string keys = param_keys + backend_keys + target_keys;
float confThreshold, nmsThreshold, scale, paddingValue;
vector<string> labels;
Scalar meanv;
bool swapRB;
int inpWidth, inpHeight;
size_t asyncNumReq = 0;
ImagePaddingMode paddingMode;
string modelName, framework;
static void preprocess(const Mat& frame, Net& net, Size inpSize);
static void postprocess(Mat& frame, const vector<Mat>& outs, Net& net, int backend, vector<int>& classIds, vector<float>& confidences, vector<Rect>& boxes, const string yolo_name);
static void drawPred(vector<int>& classIds, vector<float>& confidences, vector<Rect>& boxes, Mat& frame, FontFace& sans, int stdSize, int stdWeight, int stdImgSize, int stdThickness);
static void callback(int pos, void* userdata);
static Scalar getColor(int classId);
static void yoloPostProcessing(
const vector<Mat>& outs,
vector<int>& keep_classIds,
vector<float>& keep_confidences,
vector<Rect2d>& keep_boxes,
float conf_threshold,
float iou_threshold,
const string& yolo_name);
static void printAliases(string& zooFile){
vector<string> aliases = findAliases(zooFile, "object_detection");
cout<<"Alias choices: [ ";
for (auto it: aliases){
cout<<"'"<<it<<"' ";
}
cout<<"]"<<endl;
}
static Scalar getTextColor(Scalar bgColor) {
double luminance = 0.299 * bgColor[2] + 0.587 * bgColor[1] + 0.114 * bgColor[0];
return luminance > 128 ? Scalar(0, 0, 0) : Scalar(255, 255, 255);
}
#ifdef USE_THREADS
template <typename T>
class QueueFPS : public std::queue<T>
{
@ -112,233 +158,362 @@ private:
TickMeter tm;
std::mutex mutex;
};
#endif // USE_THREADS
int main(int argc, char** argv)
{
CommandLineParser parser(argc, argv, keys);
const std::string modelName = parser.get<String>("@alias");
const std::string zooFile = parser.get<String>("zoo");
string zooFile = parser.get<String>("zoo");
if (!parser.has("@alias") || parser.has("help"))
{
cout << about << endl;
parser.printMessage();
printAliases(zooFile);
return -1;
}
zooFile = findFile(zooFile);
modelName = parser.get<String>("@alias");
keys += genPreprocArguments(modelName, zooFile);
parser = CommandLineParser(argc, argv, keys);
parser.about("Use this script to run object detection deep learning networks using OpenCV.");
if (argc == 1 || parser.has("help"))
if (!parser.has("model"))
{
parser.printMessage();
return 0;
cout << "Path to model is not provided in command line or model alias is not correct" << endl;
printAliases(zooFile);
return -1;
}
confThreshold = parser.get<float>("thr");
nmsThreshold = parser.get<float>("nms");
float scale = parser.get<float>("scale");
Scalar mean = parser.get<Scalar>("mean");
bool swapRB = parser.get<bool>("rgb");
int inpWidth = parser.get<int>("width");
int inpHeight = parser.get<int>("height");
size_t asyncNumReq = parser.get<int>("async");
CV_Assert(parser.has("model"));
std::string modelPath = findFile(parser.get<String>("model"));
std::string configPath = findFile(parser.get<String>("config"));
//![preprocess_params]
scale = parser.get<float>("scale");
meanv = parser.get<Scalar>("mean");
swapRB = parser.get<bool>("rgb");
inpWidth = parser.get<int>("width");
inpHeight = parser.get<int>("height");
int async = parser.get<int>("async");
paddingValue = parser.get<float>("padvalue");
const string yolo_name = parser.get<String>("postprocessing");
paddingMode = static_cast<ImagePaddingMode>(parser.get<int>("paddingmode"));
//![preprocess_params]
String sha1 = parser.get<String>("sha1");
const string modelPath = findModel(parser.get<String>("model"), sha1);
const string configPath = findFile(parser.get<String>("config"));
framework = modelPath.substr(modelPath.rfind('.') + 1);
// Open file with classes names.
if (parser.has("classes"))
if (parser.has("labels"))
{
std::string file = parser.get<String>("classes");
std::ifstream ifs(file.c_str());
const string file = findFile(parser.get<String>("labels"));
ifstream ifs(file.c_str());
if (!ifs.is_open())
CV_Error(Error::StsError, "File " + file + " not found");
std::string line;
while (std::getline(ifs, line))
string line;
while (getline(ifs, line))
{
classes.push_back(line);
labels.push_back(line);
}
}
// Load a model.
Net net = readNet(modelPath, configPath, parser.get<String>("framework"));
int backend = parser.get<int>("backend");
//![read_net]
Net net = readNet(modelPath, configPath);
int backend = getBackendID(parser.get<String>("backend"));
net.setPreferableBackend(backend);
net.setPreferableTarget(parser.get<int>("target"));
std::vector<String> outNames = net.getUnconnectedOutLayersNames();
net.setPreferableTarget(getTargetID(parser.get<String>("target")));
//![read_net]
// Create a window
static const std::string kWinName = "Deep learning object detection in OpenCV";
namedWindow(kWinName, WINDOW_NORMAL);
static const string kWinName = "Deep learning object detection in OpenCV";
namedWindow(kWinName, WINDOW_AUTOSIZE);
int initialConf = (int)(confThreshold * 100);
createTrackbar("Confidence threshold, %", kWinName, &initialConf, 99, callback);
createTrackbar("Confidence threshold, %", kWinName, &initialConf, 99, callback, &net);
// Open a video file or an image file or a camera stream.
VideoCapture cap;
if (parser.has("input"))
cap.open(parser.get<String>("input"));
else
cap.open(parser.get<int>("device"));
bool openSuccess = parser.has("input") ? cap.open(parser.get<String>("input")) : cap.open(parser.get<int>("device"));
if (!openSuccess){
cout << "Could not open input file or camera device" << endl;
return 0;
}
#ifdef USE_THREADS
bool process = true;
FontFace sans("sans");
// Frames capturing thread
QueueFPS<Mat> framesQueue;
std::thread framesThread([&](){
Mat frame;
while (process)
{
cap >> frame;
if (!frame.empty())
framesQueue.push(frame.clone());
else
break;
}
});
int stdSize = 15;
int stdWeight = 150;
int stdImgSize = 512;
int stdThickness = 2;
vector<int> classIds;
vector<float> confidences;
vector<Rect> boxes;
// Frames processing thread
QueueFPS<Mat> processedFramesQueue;
QueueFPS<std::vector<Mat> > predictionsQueue;
std::thread processingThread([&](){
std::queue<AsyncArray> futureOutputs;
Mat blob;
while (process)
{
// Get a next frame
if (async > 0 && backend == DNN_BACKEND_INFERENCE_ENGINE){
asyncNumReq = async;
}
if (async != 0) {
// Threading is enabled
bool process = true;
// Frames capturing thread
QueueFPS<Mat> framesQueue;
std::thread framesThread([&]() {
Mat frame;
{
if (!framesQueue.empty())
{
frame = framesQueue.get();
if (asyncNumReq)
{
if (futureOutputs.size() == asyncNumReq)
frame = Mat();
}
else
framesQueue.clear(); // Skip the rest of frames
}
}
// Process the frame
if (!frame.empty())
{
preprocess(frame, net, Size(inpWidth, inpHeight), scale, mean, swapRB);
processedFramesQueue.push(frame);
if (asyncNumReq)
{
futureOutputs.push(net.forwardAsync());
}
while (process) {
cap >> frame;
if (!frame.empty())
framesQueue.push(frame.clone());
else
break;
}
});
// Frames processing thread
QueueFPS<Mat> processedFramesQueue;
QueueFPS<std::vector<Mat>> predictionsQueue;
std::thread processingThread([&]() {
std::queue<AsyncArray> futureOutputs;
Mat blob;
while (process) {
// Get the next frame
Mat frame;
{
std::vector<Mat> outs;
net.forward(outs, outNames);
predictionsQueue.push(outs);
if (!framesQueue.empty()) {
frame = framesQueue.get();
if (asyncNumReq) {
if (futureOutputs.size() == asyncNumReq)
frame = Mat();
}
}
}
// Process the frame
if (!frame.empty()) {
preprocess(frame, net, Size(inpWidth, inpHeight));
processedFramesQueue.push(frame);
if (asyncNumReq) {
futureOutputs.push(net.forwardAsync());
} else {
vector<Mat> outs;
net.forward(outs, net.getUnconnectedOutLayersNames());
predictionsQueue.push(outs);
}
}
while (!futureOutputs.empty() &&
futureOutputs.front().wait_for(std::chrono::seconds(0))) {
AsyncArray async_out = futureOutputs.front();
futureOutputs.pop();
Mat out;
async_out.get(out);
predictionsQueue.push({out});
}
}
});
while (!futureOutputs.empty() &&
futureOutputs.front().wait_for(std::chrono::seconds(0)))
{
AsyncArray async_out = futureOutputs.front();
futureOutputs.pop();
Mat out;
async_out.get(out);
predictionsQueue.push({out});
// Postprocessing and rendering loop
while (waitKey(100) < 0) {
if (predictionsQueue.empty())
continue;
vector<Mat> outs = predictionsQueue.get();
Mat frame = processedFramesQueue.get();
classIds.clear();
confidences.clear();
boxes.clear();
postprocess(frame, outs, net, backend, classIds, confidences, boxes, yolo_name);
drawPred(classIds, confidences, boxes, frame, sans, stdSize, stdWeight, stdImgSize, stdThickness);
int imgWidth = max(frame.rows, frame.cols);
int size = static_cast<int>((stdSize * imgWidth) / (stdImgSize * 1.5));
int weight = static_cast<int>((stdWeight * imgWidth) / (stdImgSize * 1.5));
if (predictionsQueue.counter > 1) {
string label = format("Camera: %.2f FPS", framesQueue.getFPS());
rectangle(frame, Point(0, 0), Point(10 * size, 3 * size + size / 4), Scalar::all(255), FILLED);
putText(frame, label, Point(0, size), Scalar::all(0), sans, size, weight);
label = format("Network: %.2f FPS", predictionsQueue.getFPS());
putText(frame, label, Point(0, 2 * size), Scalar::all(0), sans, size, weight);
label = format("Skipped frames: %d", framesQueue.counter - predictionsQueue.counter);
putText(frame, label, Point(0, 3 * size), Scalar::all(0), sans, size, weight);
}
imshow(kWinName, frame);
}
});
// Postprocessing and rendering loop
while (waitKey(1) < 0)
{
if (predictionsQueue.empty())
continue;
process = false;
framesThread.join();
processingThread.join();
} else {
if (asyncNumReq)
CV_Error(Error::StsNotImplemented, "Asynchronous forward is supported only with Inference Engine backend.");
// Threading is disabled, run synchronously
Mat frame, blob;
while (waitKey(100) < 0) {
cap >> frame;
if (frame.empty()) {
waitKey();
break;
}
preprocess(frame, net, Size(inpWidth, inpHeight));
std::vector<Mat> outs = predictionsQueue.get();
Mat frame = processedFramesQueue.get();
vector<Mat> outs;
net.forward(outs, net.getUnconnectedOutLayersNames());
postprocess(frame, outs, net, backend);
classIds.clear();
confidences.clear();
boxes.clear();
if (predictionsQueue.counter > 1)
{
std::string label = format("Camera: %.2f FPS", framesQueue.getFPS());
putText(frame, label, Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
postprocess(frame, outs, net, backend, classIds, confidences, boxes, yolo_name);
label = format("Network: %.2f FPS", predictionsQueue.getFPS());
putText(frame, label, Point(0, 30), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
drawPred(classIds, confidences, boxes, frame, sans, stdSize, stdWeight, stdImgSize, stdThickness);
label = format("Skipped frames: %d", framesQueue.counter - predictionsQueue.counter);
putText(frame, label, Point(0, 45), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
vector<double> layersTimes;
int imgWidth = max(frame.rows, frame.cols);
int size = static_cast<int>((stdSize * imgWidth) / (stdImgSize * 1.5));
int weight = static_cast<int>((stdWeight * imgWidth) / (stdImgSize * 1.5));
double freq = getTickFrequency() / 1000;
double t = net.getPerfProfile(layersTimes) / freq;
string label = format("Inference time: %.2f ms", t);
putText(frame, label, Point(0, size), Scalar(0, 255, 0), sans, size, weight);
imshow(kWinName, frame);
}
imshow(kWinName, frame);
}
process = false;
framesThread.join();
processingThread.join();
#else // USE_THREADS
if (asyncNumReq)
CV_Error(Error::StsNotImplemented, "Asynchronous forward is supported only with Inference Engine backend.");
// Process frames.
Mat frame, blob;
while (waitKey(1) < 0)
{
cap >> frame;
if (frame.empty())
{
waitKey();
break;
}
preprocess(frame, net, Size(inpWidth, inpHeight), scale, mean, swapRB);
std::vector<Mat> outs;
net.forward(outs, outNames);
postprocess(frame, outs, net, backend);
// Put efficiency information.
std::vector<double> layersTimes;
double freq = getTickFrequency() / 1000;
double t = net.getPerfProfile(layersTimes) / freq;
std::string label = format("Inference time: %.2f ms", t);
putText(frame, label, Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));
imshow(kWinName, frame);
}
#endif // USE_THREADS
return 0;
}
inline void preprocess(const Mat& frame, Net& net, Size inpSize, float scale,
const Scalar& mean, bool swapRB)
void preprocess(const Mat& frame, Net& net, Size inpSize)
{
static Mat blob;
// Create a 4D blob from a frame.
if (inpSize.width <= 0) inpSize.width = frame.cols;
if (inpSize.height <= 0) inpSize.height = frame.rows;
blobFromImage(frame, blob, 1.0, inpSize, Scalar(), swapRB, false, CV_8U);
Size size(inpSize.width <= 0 ? frame.cols : inpSize.width, inpSize.height <= 0 ? frame.rows : inpSize.height);
// Run a model.
net.setInput(blob, "", scale, mean);
if (net.getLayer(0)->outputNameToIndex("im_info") != -1) // Faster-RCNN or R-FCN
// Prepare the blob from the image
Mat inp;
if(framework == "weights"){ // checks whether model is darknet
blobFromImage(frame, inp, scale, size, meanv, swapRB, false, CV_32F);
}
else{
//![preprocess_call]
Image2BlobParams imgParams(
scale,
size,
meanv,
swapRB,
CV_32F,
DNN_LAYOUT_NCHW,
paddingMode,
paddingValue);
inp = blobFromImageWithParams(frame, imgParams);
//![preprocess_call]
}
// Set the blob as the network input
net.setInput(inp);
// Check if the model is Faster-RCNN or R-FCN
if (net.getLayer(0)->outputNameToIndex("im_info") != -1)
{
resize(frame, frame, inpSize);
Mat imInfo = (Mat_<float>(1, 3) << inpSize.height, inpSize.width, 1.6f);
// Resize the frame and prepare imInfo
resize(frame, frame, size);
Mat imInfo = (Mat_<float>(1, 3) << size.height, size.width, 1.6f);
net.setInput(imInfo, "im_info");
}
}
void postprocess(Mat& frame, const std::vector<Mat>& outs, Net& net, int backend)
void yoloPostProcessing(
const vector<Mat>& outs,
vector<int>& keep_classIds,
vector<float>& keep_confidences,
vector<Rect2d>& keep_boxes,
float conf_threshold,
float iou_threshold,
const string& yolo_name)
{
static std::vector<int> outLayers = net.getUnconnectedOutLayers();
static std::string outLayerType = net.getLayer(outLayers[0])->type;
// Retrieve
vector<int> classIds;
vector<float> confidences;
vector<Rect2d> boxes;
vector<Mat> outs_copy = outs;
if (yolo_name == "yolov8")
{
transposeND(outs_copy[0], {0, 2, 1}, outs_copy[0]);
}
if (yolo_name == "yolonas")
{
// outs contains 2 elements of shape [1, 8400, 80] and [1, 8400, 4]. Concat them to get [1, 8400, 84]
Mat concat_out;
// squeeze the first dimension
outs_copy[0] = outs_copy[0].reshape(1, outs_copy[0].size[1]);
outs_copy[1] = outs_copy[1].reshape(1, outs_copy[1].size[1]);
hconcat(outs_copy[1], outs_copy[0], concat_out);
outs_copy[0] = concat_out;
// remove the second element
outs_copy.pop_back();
// unsqueeze the first dimension
outs_copy[0] = outs_copy[0].reshape(0, vector<int>{1, 8400, 84});
}
for (auto preds : outs_copy)
{
preds = preds.reshape(1, preds.size[1]); // [1, 8400, 85] -> [8400, 85]
for (int i = 0; i < preds.rows; ++i)
{
// filter out non-object
float obj_conf = (yolo_name == "yolov8" || yolo_name == "yolonas") ? 1.0f : preds.at<float>(i, 4);
if (obj_conf < conf_threshold)
continue;
Mat scores = preds.row(i).colRange((yolo_name == "yolov8" || yolo_name == "yolonas") ? 4 : 5, preds.cols);
double conf;
Point maxLoc;
minMaxLoc(scores, 0, &conf, 0, &maxLoc);
conf = (yolo_name == "yolov8" || yolo_name == "yolonas") ? conf : conf * obj_conf;
if (conf < conf_threshold)
continue;
// get bbox coords
float* det = preds.ptr<float>(i);
double cx = det[0];
double cy = det[1];
double w = det[2];
double h = det[3];
// [x1, y1, x2, y2]
if (yolo_name == "yolonas") {
boxes.push_back(Rect2d(cx, cy, w, h));
} else {
boxes.push_back(Rect2d(cx - 0.5 * w, cy - 0.5 * h,
cx + 0.5 * w, cy + 0.5 * h));
}
classIds.push_back(maxLoc.x);
confidences.push_back(static_cast<float>(conf));
}
}
// NMS
vector<int> keep_idx;
NMSBoxes(boxes, confidences, conf_threshold, iou_threshold, keep_idx);
for (auto i : keep_idx)
{
keep_classIds.push_back(classIds[i]);
keep_confidences.push_back(confidences[i]);
keep_boxes.push_back(boxes[i]);
}
}
void postprocess(Mat& frame, const vector<Mat>& outs, Net& net, int backend, vector<int>& classIds, vector<float>& confidences, vector<Rect>& boxes, const string yolo_name)
{
static vector<int> outLayers = net.getUnconnectedOutLayers();
static string outLayerType = net.getLayer(outLayers[0])->type;
std::vector<int> classIds;
std::vector<float> confidences;
std::vector<Rect> boxes;
if (outLayerType == "DetectionOutput")
{
// Network produces output blob with a shape 1x1xNx7 where N is a number of
@ -405,14 +580,46 @@ void postprocess(Mat& frame, const std::vector<Mat>& outs, Net& net, int backend
}
}
}
else
CV_Error(Error::StsNotImplemented, "Unknown output layer type: " + outLayerType);
else if (outLayerType == "Identity")
{
//![forward_buffers]
vector<int> keep_classIds;
vector<float> keep_confidences;
vector<Rect2d> keep_boxes;
//![forward_buffers]
// NMS is used inside Region layer only on DNN_BACKEND_OPENCV for another backends we need NMS in sample
// or NMS is required if number of outputs > 1
//![postprocess]
yoloPostProcessing(outs, keep_classIds, keep_confidences, keep_boxes, confThreshold, nmsThreshold, yolo_name);
//![postprocess]
for (size_t i = 0; i < keep_classIds.size(); ++i)
{
classIds.push_back(keep_classIds[i]);
confidences.push_back(keep_confidences[i]);
Rect2d box = keep_boxes[i];
boxes.push_back(Rect(cvFloor(box.x), cvFloor(box.y), cvFloor(box.width-box.x), cvFloor(box.height-box.y)));
}
if (framework == "onnx"){
Image2BlobParams paramNet;
paramNet.scalefactor = scale;
paramNet.size = Size(inpWidth, inpHeight);
paramNet.mean = meanv;
paramNet.swapRB = swapRB;
paramNet.paddingmode = paddingMode;
paramNet.blobRectsToImageRects(boxes, boxes, frame.size());
}
}
else
{
CV_Error(Error::StsNotImplemented, "Unknown output layer type: " + outLayerType);
}
// NMS is used inside Region layer only on DNN_BACKEND_OPENCV for other backends we need NMS in sample
// or NMS is required if the number of outputs > 1
if (outLayers.size() > 1 || (outLayerType == "Region" && backend != DNN_BACKEND_OPENCV))
{
std::map<int, std::vector<size_t> > class2indices;
map<int, vector<size_t> > class2indices;
for (size_t i = 0; i < classIds.size(); i++)
{
if (confidences[i] >= confThreshold)
@ -420,20 +627,20 @@ void postprocess(Mat& frame, const std::vector<Mat>& outs, Net& net, int backend
class2indices[classIds[i]].push_back(i);
}
}
std::vector<Rect> nmsBoxes;
std::vector<float> nmsConfidences;
std::vector<int> nmsClassIds;
for (std::map<int, std::vector<size_t> >::iterator it = class2indices.begin(); it != class2indices.end(); ++it)
vector<Rect> nmsBoxes;
vector<float> nmsConfidences;
vector<int> nmsClassIds;
for (map<int, vector<size_t> >::iterator it = class2indices.begin(); it != class2indices.end(); ++it)
{
std::vector<Rect> localBoxes;
std::vector<float> localConfidences;
std::vector<size_t> classIndices = it->second;
vector<Rect> localBoxes;
vector<float> localConfidences;
vector<size_t> classIndices = it->second;
for (size_t i = 0; i < classIndices.size(); i++)
{
localBoxes.push_back(boxes[classIndices[i]]);
localConfidences.push_back(confidences[classIndices[i]]);
}
std::vector<int> nmsIndices;
vector<int> nmsIndices;
NMSBoxes(localBoxes, localConfidences, confThreshold, nmsThreshold, nmsIndices);
for (size_t i = 0; i < nmsIndices.size(); i++)
{
@ -447,36 +654,49 @@ void postprocess(Mat& frame, const std::vector<Mat>& outs, Net& net, int backend
classIds = nmsClassIds;
confidences = nmsConfidences;
}
for (size_t idx = 0; idx < boxes.size(); ++idx)
{
Rect box = boxes[idx];
drawPred(classIds[idx], confidences[idx], box.x, box.y,
box.x + box.width, box.y + box.height, frame);
}
}
void drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame)
void drawPred(vector<int>& classIds, vector<float>& confidences, vector<Rect>& boxes, Mat& frame, FontFace& sans, int stdSize, int stdWeight, int stdImgSize, int stdThickness)
{
rectangle(frame, Point(left, top), Point(right, bottom), Scalar(0, 255, 0));
int imgWidth = max(frame.rows, frame.cols);
int size = (stdSize*imgWidth)/stdImgSize;
int weight = (stdWeight*imgWidth)/stdImgSize;
int thickness = (stdThickness*imgWidth)/stdImgSize;
std::string label = format("%.2f", conf);
if (!classes.empty())
{
CV_Assert(classId < (int)classes.size());
label = classes[classId] + ": " + label;
for (size_t idx = 0; idx < boxes.size(); ++idx){
Scalar boxColor = getColor(classIds[idx]);
int left = boxes[idx].x;
int top = boxes[idx].y;
int right = boxes[idx].x + boxes[idx].width;
int bottom = boxes[idx].y + boxes[idx].height;
rectangle(frame, Point(left, top), Point(right, bottom), boxColor, thickness);
string label = format("%.2f", confidences[idx]);
if (!labels.empty())
{
CV_Assert(classIds[idx] < (int)labels.size());
label = labels[classIds[idx]] + ": " + label;
}
Rect r = getTextSize(Size(), label, Point(), sans, size, weight);
int baseline = r.y + r.height;
Size labelSize = Size(r.width, r.height + size/4 - baseline);
top = max(top-thickness/2, labelSize.height);
rectangle(frame, Point(left-thickness/2, top-(labelSize.height)),
Point(left + labelSize.width, top), boxColor, FILLED);
putText(frame, label, Point(left, top-size/4), getTextColor(boxColor), sans, size, weight);
}
int baseLine;
Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
top = max(top, labelSize.height);
rectangle(frame, Point(left, top - labelSize.height),
Point(left + labelSize.width, top + baseLine), Scalar::all(255), FILLED);
putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.5, Scalar());
}
void callback(int pos, void*)
{
confThreshold = pos * 0.01f;
}
Scalar getColor(int classId) {
int r = min((classId >> 0 & 1) * 128 + (classId >> 3 & 1) * 64 + (classId >> 6 & 1) * 32 + 80, 255);
int g = min((classId >> 1 & 1) * 128 + (classId >> 4 & 1) * 64 + (classId >> 7 & 1) * 32 + 40, 255);
int b = min((classId >> 2 & 1) * 128 + (classId >> 5 & 1) * 64 + (classId >> 8 & 1) * 32 + 40, 255);
return Scalar(b, g, r);
}

231
samples/dnn/object_detection.py
View File

@ -12,10 +12,22 @@ from tf_text_graph_common import readTextMessage
from tf_text_graph_ssd import createSSDGraph
from tf_text_graph_faster_rcnn import createFasterRCNNGraph
backends = (cv.dnn.DNN_BACKEND_DEFAULT, cv.dnn.DNN_BACKEND_INFERENCE_ENGINE, cv.dnn.DNN_BACKEND_OPENCV,
cv.dnn.DNN_BACKEND_VKCOM, cv.dnn.DNN_BACKEND_CUDA)
targets = (cv.dnn.DNN_TARGET_CPU, cv.dnn.DNN_TARGET_OPENCL, cv.dnn.DNN_TARGET_OPENCL_FP16, cv.dnn.DNN_TARGET_MYRIAD, cv.dnn.DNN_TARGET_HDDL,
cv.dnn.DNN_TARGET_VULKAN, cv.dnn.DNN_TARGET_CUDA, cv.dnn.DNN_TARGET_CUDA_FP16)
def help():
print(
'''
Firstly, download required models using `download_models.py` (if not already done). Set environment variable OPENCV_DOWNLOAD_CACHE_DIR to specify where models should be downloaded. Also, point OPENCV_SAMPLES_DATA_PATH to opencv/samples/data.\n"\n
To run:
python object_detection.py model_name(e.g yolov8) --input=path/to/your/input/image/or/video (don't pass --input to use device camera)
Sample command:
python object_detection.py yolov8 --input=path/to/image
Model path can also be specified using --model argument
'''
)
backends = ("default", "openvino", "opencv", "vkcom", "cuda")
targets = ("cpu", "opencl", "opencl_fp16", "ncs2_vpu", "hddl_vpu", "vulkan", "cuda", "cuda_fp16")
parser = argparse.ArgumentParser(add_help=False)
parser.add_argument('--zoo', default=os.path.join(os.path.dirname(os.path.abspath(__file__)), 'models.yml'),
@ -30,27 +42,27 @@ parser.add_argument('--framework', choices=['caffe', 'tensorflow', 'darknet', 'd
'Detect it automatically if it does not set.')
parser.add_argument('--thr', type=float, default=0.5, help='Confidence threshold')
parser.add_argument('--nms', type=float, default=0.4, help='Non-maximum suppression threshold')
parser.add_argument('--backend', choices=backends, default=cv.dnn.DNN_BACKEND_DEFAULT, type=int,
parser.add_argument('--backend', default="default", type=str, choices=backends,
help="Choose one of computation backends: "
"%d: automatically (by default), "
"%d: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit), "
"%d: OpenCV implementation, "
"%d: VKCOM, "
"%d: CUDA" % backends)
parser.add_argument('--target', choices=targets, default=cv.dnn.DNN_TARGET_CPU, type=int,
help='Choose one of target computation devices: '
'%d: CPU target (by default), '
'%d: OpenCL, '
'%d: OpenCL fp16 (half-float precision), '
'%d: NCS2 VPU, '
'%d: HDDL VPU, '
'%d: Vulkan, '
'%d: CUDA, '
'%d: CUDA fp16 (half-float preprocess)' % targets)
"default: automatically (by default), "
"openvino: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit), "
"opencv: OpenCV implementation, "
"vkcom: VKCOM, "
"cuda: CUDA, "
"webnn: WebNN")
parser.add_argument('--target', default="cpu", type=str, choices=targets,
help="Choose one of target computation devices: "
"cpu: CPU target (by default), "
"opencl: OpenCL, "
"opencl_fp16: OpenCL fp16 (half-float precision), "
"ncs2_vpu: NCS2 VPU, "
"hddl_vpu: HDDL VPU, "
"vulkan: Vulkan, "
"cuda: CUDA, "
"cuda_fp16: CUDA fp16 (half-float preprocess)")
parser.add_argument('--async', type=int, default=0,
dest='asyncN',
help='Number of asynchronous forwards at the same time. '
'Choose 0 for synchronous mode')
dest='use_threads',
help='Choose 0 for synchronous mode and 1 for asynchronous mode')
args, _ = parser.parse_known_args()
add_preproc_args(args.zoo, parser, 'object_detection')
parser = argparse.ArgumentParser(parents=[parser],
@ -58,9 +70,14 @@ parser = argparse.ArgumentParser(parents=[parser],
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
args = parser.parse_args()
args.model = findFile(args.model)
args.config = findFile(args.config)
args.classes = findFile(args.classes)
if args.alias is None or hasattr(args, 'help'):
help()
exit(1)
args.model = findModel(args.model, args.sha1)
if args.config is not None:
args.config = findFile(args.config)
args.labels = findFile(args.labels)
# If config specified, try to load it as TensorFlow Object Detection API's pipeline.
config = readTextMessage(args.config)
@ -77,40 +94,38 @@ if 'model' in config:
# Load names of classes
classes = None
if args.classes:
with open(args.classes, 'rt') as f:
classes = f.read().rstrip('\n').split('\n')
labels = None
if args.labels:
with open(args.labels, 'rt') as f:
labels = f.read().rstrip('\n').split('\n')
# Load a network
net = cv.dnn.readNet(args.model, args.config, args.framework)
net.setPreferableBackend(args.backend)
net.setPreferableTarget(args.target)
net.setPreferableBackend(get_backend_id(args.backend))
net.setPreferableTarget(get_target_id(args.target))
outNames = net.getUnconnectedOutLayersNames()
confThreshold = args.thr
nmsThreshold = args.nms
stdSize = 0.8
stdWeight = 2
stdImgSize = 512
asyncN = 0
def get_color(class_id):
r = min((class_id >> 0 & 1) * 128 + (class_id >> 3 & 1) * 64 + (class_id >> 6 & 1) * 32 + 80, 255)
g = min((class_id >> 1 & 1) * 128 + (class_id >> 4 & 1) * 64 + (class_id >> 7 & 1) * 32 + 40, 255)
b = min((class_id >> 2 & 1) * 128 + (class_id >> 5 & 1) * 64 + (class_id >> 8 & 1) * 32 + 40, 255)
return (int(b), int(g), int(r))
def get_text_color(bg_color):
luminance = 0.299 * bg_color[2] + 0.587 * bg_color[1] + 0.114 * bg_color[0]
return (0, 0, 0) if luminance > 128 else (255, 255, 255)
def postprocess(frame, outs):
frameHeight = frame.shape[0]
frameWidth = frame.shape[1]
def drawPred(classId, conf, left, top, right, bottom):
# Draw a bounding box.
cv.rectangle(frame, (left, top), (right, bottom), (0, 255, 0))
label = '%.2f' % conf
# Print a label of class.
if classes:
assert(classId < len(classes))
label = '%s: %s' % (classes[classId], label)
labelSize, baseLine = cv.getTextSize(label, cv.FONT_HERSHEY_SIMPLEX, 0.5, 1)
top = max(top, labelSize[1])
cv.rectangle(frame, (left, top - labelSize[1]), (left + labelSize[0], top + baseLine), (255, 255, 255), cv.FILLED)
cv.putText(frame, label, (left, top), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0))
layerNames = net.getLayerNames()
lastLayerId = net.getLayerId(layerNames[-1])
lastLayer = net.getLayer(lastLayerId)
@ -194,17 +209,33 @@ def postprocess(frame, outs):
else:
indices = np.arange(0, len(classIds))
return boxes, classIds, confidences, indices
def drawPred(classIds, confidences, boxes, indices, fontSize, fontThickness):
for i in indices:
box = boxes[i]
left = box[0]
top = box[1]
width = box[2]
height = box[3]
drawPred(classIds[i], confidences[i], left, top, left + width, top + height)
right = box[0] + box[2]
bottom = box[1] + box[3]
bg_color = get_color(classIds[i])
cv.rectangle(frame, (left, top), (right, bottom), bg_color, fontThickness)
label = '%.2f' % confidences[i]
# Print a label of class.
if labels:
assert(classIds[i] < len(labels))
label = '%s: %s' % (labels[classIds[i]], label)
labelSize, baseLine = cv.getTextSize(label, cv.FONT_HERSHEY_SIMPLEX, fontSize, fontThickness)
top = max(top, labelSize[1])
cv.rectangle(frame, (int(left-fontThickness/2), top - labelSize[1]), (left + labelSize[0], top + baseLine), bg_color, cv.FILLED)
cv.putText(frame, label, (left, top-fontThickness), cv.FONT_HERSHEY_SIMPLEX, fontSize, get_text_color(bg_color), fontThickness)
# Process inputs
winName = 'Deep learning object detection in OpenCV'
cv.namedWindow(winName, cv.WINDOW_NORMAL)
cv.namedWindow(winName, cv.WINDOW_AUTOSIZE)
def callback(pos):
global confThreshold
@ -252,7 +283,7 @@ def framesThreadBody():
processedFramesQueue = queue.Queue()
predictionsQueue = QueueFPS()
def processingThreadBody():
global processedFramesQueue, predictionsQueue, args, process
global processedFramesQueue, predictionsQueue, args, process, asyncN
futureOutputs = []
while process:
@ -261,8 +292,8 @@ def processingThreadBody():
try:
frame = framesQueue.get_nowait()
if args.asyncN:
if len(futureOutputs) == args.asyncN:
if asyncN:
if len(futureOutputs) == asyncN:
frame = None # Skip the frame
else:
framesQueue.queue.clear() # Skip the rest of frames
@ -277,7 +308,7 @@ def processingThreadBody():
# Create a 4D blob from a frame.
inpWidth = args.width if args.width else frameWidth
inpHeight = args.height if args.height else frameHeight
blob = cv.dnn.blobFromImage(frame, size=(inpWidth, inpHeight), swapRB=args.rgb, ddepth=cv.CV_8U)
blob = cv.dnn.blobFromImage(frame, size=(inpWidth, inpHeight), swapRB=args.rgb, ddepth=cv.CV_32F)
processedFramesQueue.put(frame)
# Run a model
@ -286,7 +317,7 @@ def processingThreadBody():
frame = cv.resize(frame, (inpWidth, inpHeight))
net.setInput(np.array([[inpHeight, inpWidth, 1.6]], dtype=np.float32), 'im_info')
if args.asyncN:
if asyncN:
futureOutputs.append(net.forwardAsync())
else:
outs = net.forward(outNames)
@ -298,40 +329,68 @@ def processingThreadBody():
del futureOutputs[0]
if args.use_threads:
framesThread = Thread(target=framesThreadBody)
framesThread.start()
framesThread = Thread(target=framesThreadBody)
framesThread.start()
processingThread = Thread(target=processingThreadBody)
processingThread.start()
processingThread = Thread(target=processingThreadBody)
processingThread.start()
#
# Postprocessing and rendering loop
#
while cv.waitKey(1) < 0:
try:
# Request prediction first because they put after frames
outs = predictionsQueue.get_nowait()
frame = processedFramesQueue.get_nowait()
imgWidth = max(frame.shape[:2])
fontSize = (stdSize*imgWidth)/stdImgSize
fontThickness = max(1,(stdWeight*imgWidth)//stdImgSize)
#
# Postprocessing and rendering loop
#
while cv.waitKey(1) < 0:
try:
# Request prediction first because they put after frames
outs = predictionsQueue.get_nowait()
frame = processedFramesQueue.get_nowait()
boxes, classIds, confidences, indices = postprocess(frame, outs)
drawPred(classIds, confidences, boxes, indices, fontSize, fontThickness)
fontSize = fontSize/2
# Put efficiency information.
if predictionsQueue.counter > 1:
label = 'Camera: %.2f FPS' % (framesQueue.getFPS())
cv.rectangle(frame, (0, 0), (int(260*fontSize), int(80*fontSize)), (255,255,255), cv.FILLED)
cv.putText(frame, label, (0, int(25*fontSize)), cv.FONT_HERSHEY_SIMPLEX, fontSize, (0, 0, 0), fontThickness)
postprocess(frame, outs)
label = 'Network: %.2f FPS' % (predictionsQueue.getFPS())
cv.putText(frame, label, (0, int(2*25*fontSize)), cv.FONT_HERSHEY_SIMPLEX, fontSize, (0, 0, 0), fontThickness)
# Put efficiency information.
if predictionsQueue.counter > 1:
label = 'Camera: %.2f FPS' % (framesQueue.getFPS())
cv.putText(frame, label, (0, 15), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0))
label = 'Skipped frames: %d' % (framesQueue.counter - predictionsQueue.counter)
cv.putText(frame, label, (0, int(3*25*fontSize)), cv.FONT_HERSHEY_SIMPLEX, fontSize, (0, 0, 0), fontThickness)
label = 'Network: %.2f FPS' % (predictionsQueue.getFPS())
cv.putText(frame, label, (0, 30), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0))
label = 'Skipped frames: %d' % (framesQueue.counter - predictionsQueue.counter)
cv.putText(frame, label, (0, 45), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0))
cv.imshow(winName, frame)
except queue.Empty:
pass
cv.imshow(winName, frame)
except queue.Empty:
pass
process = False
framesThread.join()
processingThread.join()
process = False
framesThread.join()
processingThread.join()
else:
# Non-threaded processing if --async is 0
while cv.waitKey(1) < 0:
hasFrame, frame = cap.read()
if not hasFrame:
cv.waitKey()
break
frameHeight = frame.shape[0]
frameWidth = frame.shape[1]
inpWidth = args.width if args.width else frameWidth
inpHeight = args.height if args.height else frameHeight
blob = cv.dnn.blobFromImage(frame, size=(inpWidth, inpHeight), swapRB=args.rgb, ddepth=cv.CV_32F)
net.setInput(blob, scalefactor=args.scale, mean=args.mean)
outs = net.forward(outNames)
boxes, classIds, confidences, indices = postprocess(frame, outs)
drawPred(classIds, confidences, boxes, indices, (stdSize*max(frame.shape[:2]))/stdImgSize, (stdWeight*max(frame.shape[:2]))//stdImgSize)
cv.imshow(winName, frame)

382
samples/dnn/yolo_detector.cpp
View File

@ -1,382 +0,0 @@
/**
* @file yolo_detector.cpp
* @brief Yolo Object Detection Sample
* @author OpenCV team
*/
//![includes]
#include <opencv2/dnn.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/imgcodecs.hpp>
#include <fstream>
#include <sstream>
#include "iostream"
#include "common.hpp"
#include <opencv2/highgui.hpp>
//![includes]
using namespace cv;
using namespace cv::dnn;
void getClasses(std::string classesFile);
void drawPrediction(int classId, float conf, int left, int top, int right, int bottom, Mat& frame);
void yoloPostProcessing(
std::vector<Mat>& outs,
std::vector<int>& keep_classIds,
std::vector<float>& keep_confidences,
std::vector<Rect2d>& keep_boxes,
float conf_threshold,
float iou_threshold,
const std::string& model_name,
const int nc
);
std::vector<std::string> classes;
std::string keys =
"{ help h | | Print help message. }"
"{ device | 0 | camera device number. }"
"{ model | onnx/models/yolox_s_inf_decoder.onnx | Default model. }"
"{ yolo | yolox | yolo model version. }"
"{ input i | | Path to input image or video file. Skip this argument to capture frames from a camera. }"
"{ classes | | Optional path to a text file with names of classes to label detected objects. }"
"{ nc | 80 | Number of classes. Default is 80 (coming from COCO dataset). }"
"{ thr | .5 | Confidence threshold. }"
"{ nms | .4 | Non-maximum suppression threshold. }"
"{ mean | 0.0 | Normalization constant. }"
"{ scale | 1.0 | Preprocess input image by multiplying on a scale factor. }"
"{ width | 640 | Preprocess input image by resizing to a specific width. }"
"{ height | 640 | Preprocess input image by resizing to a specific height. }"
"{ rgb | 1 | Indicate that model works with RGB input images instead BGR ones. }"
"{ padvalue | 114.0 | padding value. }"
"{ paddingmode | 2 | Choose one of computation backends: "
"0: resize to required input size without extra processing, "
"1: Image will be cropped after resize, "
"2: Resize image to the desired size while preserving the aspect ratio of original image }"
"{ backend | 0 | Choose one of computation backends: "
"0: automatically (by default), "
"1: Halide language (http://halide-lang.org/), "
"2: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit), "
"3: OpenCV implementation, "
"4: VKCOM, "
"5: CUDA }"
"{ target | 0 | Choose one of target computation devices: "
"0: CPU target (by default), "
"1: OpenCL, "
"2: OpenCL fp16 (half-float precision), "
"3: VPU, "
"4: Vulkan, "
"6: CUDA, "
"7: CUDA fp16 (half-float preprocess) }"
"{ async | 0 | Number of asynchronous forwards at the same time. "
"Choose 0 for synchronous mode }";
void getClasses(std::string classesFile)
{
std::ifstream ifs(classesFile.c_str());
if (!ifs.is_open())
CV_Error(Error::StsError, "File " + classesFile + " not found");
std::string line;
while (std::getline(ifs, line))
classes.push_back(line);
}
void drawPrediction(int classId, float conf, int left, int top, int right, int bottom, Mat& frame)
{
rectangle(frame, Point(left, top), Point(right, bottom), Scalar(0, 255, 0));
std::string label = format("%.2f", conf);
if (!classes.empty())
{
CV_Assert(classId < (int)classes.size());
label = classes[classId] + ": " + label;
}
int baseLine;
Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
top = max(top, labelSize.height);
rectangle(frame, Point(left, top - labelSize.height),
Point(left + labelSize.width, top + baseLine), Scalar::all(255), FILLED);
putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.5, Scalar());
}
void yoloPostProcessing(
std::vector<Mat>& outs,
std::vector<int>& keep_classIds,
std::vector<float>& keep_confidences,
std::vector<Rect2d>& keep_boxes,
float conf_threshold,
float iou_threshold,
const std::string& model_name,
const int nc=80)
{
// Retrieve
std::vector<int> classIds;
std::vector<float> confidences;
std::vector<Rect2d> boxes;
if (model_name == "yolov8" || model_name == "yolov10" ||
model_name == "yolov9")
{
cv::transposeND(outs[0], {0, 2, 1}, outs[0]);
}
if (model_name == "yolonas")
{
// outs contains 2 elemets of shape [1, 8400, 80] and [1, 8400, 4]. Concat them to get [1, 8400, 84]
Mat concat_out;
// squeeze the first dimension
outs[0] = outs[0].reshape(1, outs[0].size[1]);
outs[1] = outs[1].reshape(1, outs[1].size[1]);
cv::hconcat(outs[1], outs[0], concat_out);
outs[0] = concat_out;
// remove the second element
outs.pop_back();
// unsqueeze the first dimension
outs[0] = outs[0].reshape(0, std::vector<int>{1, 8400, nc + 4});
}
// assert if last dim is 85 or 84
CV_CheckEQ(outs[0].dims, 3, "Invalid output shape. The shape should be [1, #anchors, 85 or 84]");
CV_CheckEQ((outs[0].size[2] == nc + 5 || outs[0].size[2] == 80 + 4), true, "Invalid output shape: ");
for (auto preds : outs)
{
preds = preds.reshape(1, preds.size[1]); // [1, 8400, 85] -> [8400, 85]
for (int i = 0; i < preds.rows; ++i)
{
// filter out non object
float obj_conf = (model_name == "yolov8" || model_name == "yolonas" ||
model_name == "yolov9" || model_name == "yolov10") ? 1.0f : preds.at<float>(i, 4) ;
if (obj_conf < conf_threshold)
continue;
Mat scores = preds.row(i).colRange((model_name == "yolov8" || model_name == "yolonas" || model_name == "yolov9" || model_name == "yolov10") ? 4 : 5, preds.cols);
double conf;
Point maxLoc;
minMaxLoc(scores, 0, &conf, 0, &maxLoc);
conf = (model_name == "yolov8" || model_name == "yolonas" || model_name == "yolov9" || model_name == "yolov10") ? conf : conf * obj_conf;
if (conf < conf_threshold)
continue;
// get bbox coords
float* det = preds.ptr<float>(i);
double cx = det[0];
double cy = det[1];
double w = det[2];
double h = det[3];
// [x1, y1, x2, y2]
if (model_name == "yolonas" || model_name == "yolov10"){
boxes.push_back(Rect2d(cx, cy, w, h));
} else {
boxes.push_back(Rect2d(cx - 0.5 * w, cy - 0.5 * h,
cx + 0.5 * w, cy + 0.5 * h));
}
classIds.push_back(maxLoc.x);
confidences.push_back(static_cast<float>(conf));
}
}
// NMS
std::vector<int> keep_idx;
NMSBoxes(boxes, confidences, conf_threshold, iou_threshold, keep_idx);
for (auto i : keep_idx)
{
keep_classIds.push_back(classIds[i]);
keep_confidences.push_back(confidences[i]);
keep_boxes.push_back(boxes[i]);
}
}
/**
* @function main
* @brief Main function
*/
int main(int argc, char** argv)
{
CommandLineParser parser(argc, argv, keys);
parser.about("Use this script to run object detection deep learning networks using OpenCV.");
if (parser.has("help"))
{
parser.printMessage();
return 0;
}
CV_Assert(parser.has("model"));
CV_Assert(parser.has("yolo"));
// if model is default, use findFile to get the full path otherwise use the given path
std::string weightPath = findFile(parser.get<String>("model"));
std::string yolo_model = parser.get<String>("yolo");
int nc = parser.get<int>("nc");
float confThreshold = parser.get<float>("thr");
float nmsThreshold = parser.get<float>("nms");
//![preprocess_params]
float paddingValue = parser.get<float>("padvalue");
bool swapRB = parser.get<bool>("rgb");
int inpWidth = parser.get<int>("width");
int inpHeight = parser.get<int>("height");
Scalar scale = parser.get<float>("scale");
Scalar mean = parser.get<Scalar>("mean");
ImagePaddingMode paddingMode = static_cast<ImagePaddingMode>(parser.get<int>("paddingmode"));
//![preprocess_params]
// check if yolo model is valid
if (yolo_model != "yolov5" && yolo_model != "yolov6"
&& yolo_model != "yolov7" && yolo_model != "yolov8"
&& yolo_model != "yolov10" && yolo_model !="yolov9"
&& yolo_model != "yolox" && yolo_model != "yolonas")
CV_Error(Error::StsError, "Invalid yolo model: " + yolo_model);
// get classes
if (parser.has("classes"))
{
getClasses(findFile(parser.get<String>("classes")));
}
// load model
//![read_net]
Net net = readNet(weightPath);
int backend = parser.get<int>("backend");
net.setPreferableBackend(backend);
net.setPreferableTarget(parser.get<int>("target"));
//![read_net]
VideoCapture cap;
Mat img;
bool isImage = false;
bool isCamera = false;
// Check if input is given
if (parser.has("input"))
{
String input = parser.get<String>("input");
// Check if the input is an image
if (input.find(".jpg") != String::npos || input.find(".png") != String::npos)
{
img = imread(findFile(input));
if (img.empty())
{
CV_Error(Error::StsError, "Cannot read image file: " + input);
}
isImage = true;
}
else
{
cap.open(input);
if (!cap.isOpened())
{
CV_Error(Error::StsError, "Cannot open video " + input);
}
isCamera = true;
}
}
else
{
int cameraIndex = parser.get<int>("device");
cap.open(cameraIndex);
if (!cap.isOpened())
{
CV_Error(Error::StsError, cv::format("Cannot open camera #%d", cameraIndex));
}
isCamera = true;
}
// image pre-processing
//![preprocess_call]
Size size(inpWidth, inpHeight);
Image2BlobParams imgParams(
scale,
size,
mean,
swapRB,
CV_32F,
DNN_LAYOUT_NCHW,
paddingMode,
paddingValue);
// rescale boxes back to original image
Image2BlobParams paramNet;
paramNet.scalefactor = scale;
paramNet.size = size;
paramNet.mean = mean;
paramNet.swapRB = swapRB;
paramNet.paddingmode = paddingMode;
//![preprocess_call]
//![forward_buffers]
std::vector<Mat> outs;
std::vector<int> keep_classIds;
std::vector<float> keep_confidences;
std::vector<Rect2d> keep_boxes;
std::vector<Rect> boxes;
//![forward_buffers]
Mat inp;
while (waitKey(1) < 0)
{
if (isCamera)
cap >> img;
if (img.empty())
{
std::cout << "Empty frame" << std::endl;
waitKey();
break;
}
//![preprocess_call_func]
inp = blobFromImageWithParams(img, imgParams);
//![preprocess_call_func]
//![forward]
net.setInput(inp);
net.forward(outs, net.getUnconnectedOutLayersNames());
//![forward]
//![postprocess]
yoloPostProcessing(
outs, keep_classIds, keep_confidences, keep_boxes,
confThreshold, nmsThreshold,
yolo_model,
nc);
//![postprocess]
// covert Rect2d to Rect
//![draw_boxes]
for (auto box : keep_boxes)
{
boxes.push_back(Rect(cvFloor(box.x), cvFloor(box.y), cvFloor(box.width - box.x), cvFloor(box.height - box.y)));
}
paramNet.blobRectsToImageRects(boxes, boxes, img.size());
for (size_t idx = 0; idx < boxes.size(); ++idx)
{
Rect box = boxes[idx];
drawPrediction(keep_classIds[idx], keep_confidences[idx], box.x, box.y,
box.width + box.x, box.height + box.y, img);
}
const std::string kWinName = "Yolo Object Detector";
namedWindow(kWinName, WINDOW_NORMAL);
imshow(kWinName, img);
//![draw_boxes]
outs.clear();
keep_classIds.clear();
keep_confidences.clear();
keep_boxes.clear();
boxes.clear();
if (isImage)
{
waitKey();
break;
}
}
}