mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-01-18 22:44:02 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/3.4' into merge-3.4

Browse Source
This commit is contained in:
Alexander Alekhin 2021-11-27 13:17:44 +00:00
commit 57ee14d62d
37 changed files with 527 additions and 111 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
3rdparty/libjpeg-turbo/CMakeLists.txt vendored
View File

@ -4,9 +4,9 @@ ocv_warnings_disable(CMAKE_C_FLAGS -Wunused-parameter -Wsign-compare -Wshorten-6
set(VERSION_MAJOR 2)
set(VERSION_MINOR 1)
set(VERSION_REVISION 0)
set(VERSION_REVISION 2)
set(VERSION ${VERSION_MAJOR}.${VERSION_MINOR}.${VERSION_REVISION})
set(LIBJPEG_TURBO_VERSION_NUMBER 2001000)
set(LIBJPEG_TURBO_VERSION_NUMBER 2001002)
string(TIMESTAMP BUILD "opencv-${OPENCV_VERSION}-libjpeg-turbo")
if(CMAKE_BUILD_TYPE STREQUAL "Debug")

10
3rdparty/libjpeg-turbo/jconfigint.h.in vendored
View File

@ -40,3 +40,13 @@
#define HAVE_BITSCANFORWARD
#endif
#endif
#if defined(__has_attribute)
#if __has_attribute(fallthrough)
#define FALLTHROUGH __attribute__((fallthrough));
#else
#define FALLTHROUGH
#endif
#else
#define FALLTHROUGH
#endif

5
3rdparty/libjpeg-turbo/src/jchuff.c vendored
View File

@ -44,8 +44,9 @@
* flags (this defines __thumb__).
*/
#if defined(__arm__) || defined(__aarch64__) || defined(_M_ARM) || \
defined(_M_ARM64)
/* NOTE: Both GCC and Clang define __GNUC__ */
#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
defined(_M_ARM) || defined(_M_ARM64)
#if !defined(__thumb__) || defined(__thumb2__)
#define USE_CLZ_INTRINSIC
#endif

2
3rdparty/libjpeg-turbo/src/jcmaster.c vendored
View File

@ -493,7 +493,7 @@ prepare_for_pass(j_compress_ptr cinfo)
master->pass_type = output_pass;
master->pass_number++;
#endif
/*FALLTHROUGH*/
FALLTHROUGH /*FALLTHROUGH*/
case output_pass:
/* Do a data-output pass. */
/* We need not repeat per-scan setup if prior optimization pass did it. */

10
3rdparty/libjpeg-turbo/src/jcphuff.c vendored
View File

@ -7,6 +7,7 @@
* Copyright (C) 2011, 2015, 2018, 2021, D. R. Commander.
* Copyright (C) 2016, 2018, Matthieu Darbois.
* Copyright (C) 2020, Arm Limited.
* Copyright (C) 2021, Alex Richardson.
* For conditions of distribution and use, see the accompanying README.ijg
* file.
*
@ -52,8 +53,9 @@
* flags (this defines __thumb__).
*/
#if defined(__arm__) || defined(__aarch64__) || defined(_M_ARM) || \
defined(_M_ARM64)
/* NOTE: Both GCC and Clang define __GNUC__ */
#if (defined(__GNUC__) && (defined(__arm__) || defined(__aarch64__))) || \
defined(_M_ARM) || defined(_M_ARM64)
#if !defined(__thumb__) || defined(__thumb2__)
#define USE_CLZ_INTRINSIC
#endif
@ -679,7 +681,7 @@ encode_mcu_AC_first(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
emit_restart(entropy, entropy->next_restart_num);
#ifdef WITH_SIMD
cvalue = values = (JCOEF *)PAD((size_t)values_unaligned, 16);
cvalue = values = (JCOEF *)PAD((JUINTPTR)values_unaligned, 16);
#else
/* Not using SIMD, so alignment is not needed */
cvalue = values = values_unaligned;
@ -944,7 +946,7 @@ encode_mcu_AC_refine(j_compress_ptr cinfo, JBLOCKROW *MCU_data)
emit_restart(entropy, entropy->next_restart_num);
#ifdef WITH_SIMD
cabsvalue = absvalues = (JCOEF *)PAD((size_t)absvalues_unaligned, 16);
cabsvalue = absvalues = (JCOEF *)PAD((JUINTPTR)absvalues_unaligned, 16);
#else
/* Not using SIMD, so alignment is not needed */
cabsvalue = absvalues = absvalues_unaligned;

3
3rdparty/libjpeg-turbo/src/jdapimin.c vendored
View File

@ -23,6 +23,7 @@
#include "jinclude.h"
#include "jpeglib.h"
#include "jdmaster.h"
#include "jconfigint.h"
/*
@ -308,7 +309,7 @@ jpeg_consume_input(j_decompress_ptr cinfo)
/* Initialize application's data source module */
(*cinfo->src->init_source) (cinfo);
cinfo->global_state = DSTATE_INHEADER;
/*FALLTHROUGH*/
FALLTHROUGH /*FALLTHROUGH*/
case DSTATE_INHEADER:
retcode = (*cinfo->inputctl->consume_input) (cinfo);
if (retcode == JPEG_REACHED_SOS) { /* Found SOS, prepare to decompress */

11
3rdparty/libjpeg-turbo/src/jdhuff.c vendored
View File

@ -584,7 +584,7 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
* behavior is, to the best of our understanding, innocuous, and it is
* unclear how to work around it without potentially affecting
* performance. Thus, we (hopefully temporarily) suppress UBSan integer
* overflow errors for this function.
* overflow errors for this function and decode_mcu_fast().
*/
s += state.last_dc_val[ci];
state.last_dc_val[ci] = s;
@ -651,6 +651,12 @@ decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
}
#if defined(__has_feature)
#if __has_feature(undefined_behavior_sanitizer)
__attribute__((no_sanitize("signed-integer-overflow"),
no_sanitize("unsigned-integer-overflow")))
#endif
#endif
LOCAL(boolean)
decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
{
@ -681,6 +687,9 @@ decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
if (entropy->dc_needed[blkn]) {
int ci = cinfo->MCU_membership[blkn];
/* Refer to the comment in decode_mcu_slow() regarding the supression of
* a UBSan integer overflow error in this line of code.
*/
s += state.last_dc_val[ci];
state.last_dc_val[ci] = s;
if (block)

5
3rdparty/libjpeg-turbo/src/jdmainct.c vendored
View File

@ -18,6 +18,7 @@
#include "jinclude.h"
#include "jdmainct.h"
#include "jconfigint.h"
/*
@ -360,7 +361,7 @@ process_data_context_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
if (*out_row_ctr >= out_rows_avail)
return; /* Postprocessor exactly filled output buf */
/*FALLTHROUGH*/
FALLTHROUGH /*FALLTHROUGH*/
case CTX_PREPARE_FOR_IMCU:
/* Prepare to process first M-1 row groups of this iMCU row */
main_ptr->rowgroup_ctr = 0;
@ -371,7 +372,7 @@ process_data_context_main(j_decompress_ptr cinfo, JSAMPARRAY output_buf,
if (main_ptr->iMCU_row_ctr == cinfo->total_iMCU_rows)
set_bottom_pointers(cinfo);
main_ptr->context_state = CTX_PROCESS_IMCU;
/*FALLTHROUGH*/
FALLTHROUGH /*FALLTHROUGH*/
case CTX_PROCESS_IMCU:
/* Call postprocessor using previously set pointers */
(*cinfo->post->post_process_data) (cinfo,

6
3rdparty/libjpeg-turbo/src/jmemmgr.c vendored
View File

@ -4,7 +4,7 @@
* This file was part of the Independent JPEG Group's software:
* Copyright (C) 1991-1997, Thomas G. Lane.
* libjpeg-turbo Modifications:
* Copyright (C) 2016, D. R. Commander.
* Copyright (C) 2016, 2021, D. R. Commander.
* For conditions of distribution and use, see the accompanying README.ijg
* file.
*
@ -1032,7 +1032,7 @@ free_pool(j_common_ptr cinfo, int pool_id)
large_pool_ptr next_lhdr_ptr = lhdr_ptr->next;
space_freed = lhdr_ptr->bytes_used +
lhdr_ptr->bytes_left +
sizeof(large_pool_hdr);
sizeof(large_pool_hdr) + ALIGN_SIZE - 1;
jpeg_free_large(cinfo, (void *)lhdr_ptr, space_freed);
mem->total_space_allocated -= space_freed;
lhdr_ptr = next_lhdr_ptr;
@ -1045,7 +1045,7 @@ free_pool(j_common_ptr cinfo, int pool_id)
while (shdr_ptr != NULL) {
small_pool_ptr next_shdr_ptr = shdr_ptr->next;
space_freed = shdr_ptr->bytes_used + shdr_ptr->bytes_left +
sizeof(small_pool_hdr);
sizeof(small_pool_hdr) + ALIGN_SIZE - 1;
jpeg_free_small(cinfo, (void *)shdr_ptr, space_freed);
mem->total_space_allocated -= space_freed;
shdr_ptr = next_shdr_ptr;

15
3rdparty/libjpeg-turbo/src/jpegint.h vendored
View File

@ -5,8 +5,9 @@
* Copyright (C) 1991-1997, Thomas G. Lane.
* Modified 1997-2009 by Guido Vollbeding.
* libjpeg-turbo Modifications:
* Copyright (C) 2015-2016, 2019, D. R. Commander.
* Copyright (C) 2015-2016, 2019, 2021, D. R. Commander.
* Copyright (C) 2015, Google, Inc.
* Copyright (C) 2021, Alex Richardson.
* For conditions of distribution and use, see the accompanying README.ijg
* file.
*
@ -47,6 +48,18 @@ typedef enum { /* Operating modes for buffer controllers */
/* JLONG must hold at least signed 32-bit values. */
typedef long JLONG;
/* JUINTPTR must hold pointer values. */
#ifdef __UINTPTR_TYPE__
/*
* __UINTPTR_TYPE__ is GNU-specific and available in GCC 4.6+ and Clang 3.0+.
* Fortunately, that is sufficient to support the few architectures for which
* sizeof(void *) != sizeof(size_t). The only other options would require C99
* or Clang-specific builtins.
*/
typedef __UINTPTR_TYPE__ JUINTPTR;
#else
typedef size_t JUINTPTR;
#endif
/*
* Left shift macro that handles a negative operand without causing any

2
CMakeLists.txt
View File

@ -106,7 +106,7 @@ set(CMAKE_POSITION_INDEPENDENT_CODE ${ENABLE_PIC})
ocv_cmake_hook(PRE_CMAKE_BOOTSTRAP)
# Bootstap CMake system: setup CMAKE_SYSTEM_NAME and other vars
# Bootstrap CMake system: setup CMAKE_SYSTEM_NAME and other vars
if(OPENCV_WORKAROUND_CMAKE_20989)
set(CMAKE_SYSTEM_PROCESSOR_BACKUP ${CMAKE_SYSTEM_PROCESSOR})
endif()

2
doc/js_tutorials/js_assets/js_camshift.html
View File

@ -77,7 +77,7 @@ cv.normalize(roiHist, roiHist, 0, 255, cv.NORM_MINMAX);
// delete useless mats.
roi.delete(); hsvRoi.delete(); mask.delete(); low.delete(); high.delete(); hsvRoiVec.delete();
// Setup the termination criteria, either 10 iteration or move by atleast 1 pt
// Setup the termination criteria, either 10 iteration or move by at least 1 pt
let termCrit = new cv.TermCriteria(cv.TERM_CRITERIA_EPS | cv.TERM_CRITERIA_COUNT, 10, 1);
let hsv = new cv.Mat(video.height, video.width, cv.CV_8UC3);

2
doc/js_tutorials/js_assets/js_meanshift.html
View File

@ -77,7 +77,7 @@ cv.normalize(roiHist, roiHist, 0, 255, cv.NORM_MINMAX);
// delete useless mats.
roi.delete(); hsvRoi.delete(); mask.delete(); low.delete(); high.delete(); hsvRoiVec.delete();
// Setup the termination criteria, either 10 iteration or move by atleast 1 pt
// Setup the termination criteria, either 10 iteration or move by at least 1 pt
let termCrit = new cv.TermCriteria(cv.TERM_CRITERIA_EPS | cv.TERM_CRITERIA_COUNT, 10, 1);
let hsv = new cv.Mat(video.height, video.width, cv.CV_8UC3);

26
modules/core/src/logger.cpp
View File

@ -181,17 +181,33 @@ LogLevel getLogLevel()
namespace internal {
static int getShowTimestampMode()
{
static bool param_timestamp_enable = utils::getConfigurationParameterBool("OPENCV_LOG_TIMESTAMP", true);
static bool param_timestamp_ns_enable = utils::getConfigurationParameterBool("OPENCV_LOG_TIMESTAMP_NS", false);
return (param_timestamp_enable ? 1 : 0) + (param_timestamp_ns_enable ? 2 : 0);
}
void writeLogMessage(LogLevel logLevel, const char* message)
{
const int threadID = cv::utils::getThreadID();
std::string message_id;
switch (getShowTimestampMode())
{
case 1: message_id = cv::format("%d@%0.3f", threadID, getTimestampNS() * 1e-9); break;
case 1+2: message_id = cv::format("%d@%llu", threadID, (long long unsigned int)getTimestampNS()); break;
default: message_id = cv::format("%d", threadID); break;
}
std::ostringstream ss;
switch (logLevel)
{
case LOG_LEVEL_FATAL: ss << "[FATAL:" << threadID << "] " << message << std::endl; break;
case LOG_LEVEL_ERROR: ss << "[ERROR:" << threadID << "] " << message << std::endl; break;
case LOG_LEVEL_WARNING: ss << "[ WARN:" << threadID << "] " << message << std::endl; break;
case LOG_LEVEL_INFO: ss << "[ INFO:" << threadID << "] " << message << std::endl; break;
case LOG_LEVEL_DEBUG: ss << "[DEBUG:" << threadID << "] " << message << std::endl; break;
case LOG_LEVEL_FATAL: ss << "[FATAL:" << message_id << "] " << message << std::endl; break;
case LOG_LEVEL_ERROR: ss << "[ERROR:" << message_id << "] " << message << std::endl; break;
case LOG_LEVEL_WARNING: ss << "[ WARN:" << message_id << "] " << message << std::endl; break;
case LOG_LEVEL_INFO: ss << "[ INFO:" << message_id << "] " << message << std::endl; break;
case LOG_LEVEL_DEBUG: ss << "[DEBUG:" << message_id << "] " << message << std::endl; break;
case LOG_LEVEL_VERBOSE: ss << message << std::endl; break;
case LOG_LEVEL_SILENT: return; // avoid compiler warning about incomplete switch
case ENUM_LOG_LEVEL_FORCE_INT: return; // avoid compiler warning about incomplete switch

4
modules/core/src/precomp.hpp
View File

@ -368,6 +368,10 @@ bool __termination; // skip some cleanups, because process is terminating
cv::Mutex& getInitializationMutex();
/// @brief Returns timestamp in nanoseconds since program launch
int64 getTimestampNS();
#define CV_SINGLETON_LAZY_INIT_(TYPE, INITIALIZER, RET_VALUE) \
static TYPE* const instance = INITIALIZER; \
return RET_VALUE;

45
modules/core/src/system.cpp
View File

@ -944,6 +944,51 @@ int64 getCPUTickCount(void)
#endif
namespace internal {
class Timestamp
{
public:
const int64 zeroTickCount;
const double ns_in_ticks;
Timestamp()
: zeroTickCount(getTickCount())
, ns_in_ticks(1e9 / getTickFrequency())
{
// nothing
}
int64 getTimestamp()
{
int64 t = getTickCount();
return (int64)((t - zeroTickCount) * ns_in_ticks);
}
static Timestamp& getInstance()
{
static Timestamp g_timestamp;
return g_timestamp;
}
};
class InitTimestamp {
public:
InitTimestamp() {
Timestamp::getInstance();
}
};
static InitTimestamp g_initialize_timestamp; // force zero timestamp initialization
} // namespace
int64 getTimestampNS()
{
return internal::Timestamp::getInstance().getTimestamp();
}
const String& getBuildInformation()
{
static String build_info =

17
modules/core/src/trace.cpp
View File

@ -63,15 +63,6 @@ namespace details {
#pragma warning(disable:4065) // switch statement contains 'default' but no 'case' labels
#endif
static int64 g_zero_timestamp = 0;
static int64 getTimestamp()
{
int64 t = getTickCount();
static double tick_to_ns = 1e9 / getTickFrequency();
return (int64)((t - g_zero_timestamp) * tick_to_ns);
}
static bool getParameterTraceEnable()
{
static bool param_traceEnable = utils::getConfigurationParameterBool("OPENCV_TRACE", false);
@ -485,7 +476,7 @@ Region::Region(const LocationStaticStorage& location) :
}
}
int64 beginTimestamp = getTimestamp();
int64 beginTimestamp = getTimestampNS();
int currentDepth = ctx.getCurrentDepth() + 1;
switch (location.flags & REGION_FLAG_IMPL_MASK)
@ -635,7 +626,7 @@ void Region::destroy()
}
}
int64 endTimestamp = getTimestamp();
int64 endTimestamp = getTimestampNS();
int64 duration = endTimestamp - ctx.stackTopBeginTimestamp();
bool active = isActive();
@ -844,7 +835,7 @@ static bool isInitialized = false;
TraceManager::TraceManager()
{
g_zero_timestamp = cv::getTickCount();
(void)cv::getTimestampNS();
isInitialized = true;
CV_LOG("TraceManager ctor: " << (void*)this);
@ -990,7 +981,7 @@ void parallelForFinalize(const Region& rootRegion)
{
TraceManagerThreadLocal& ctx = getTraceManager().tls.getRef();
int64 endTimestamp = getTimestamp();
int64 endTimestamp = getTimestampNS();
int64 duration = endTimestamp - ctx.stackTopBeginTimestamp();
CV_LOG_PARALLEL(NULL, "parallel_for duration: " << duration << " " << &rootRegion);

14
modules/dnn/test/test_backends.cpp
View File

@ -217,8 +217,16 @@ TEST_P(DNNTestNetwork, MobileNet_SSD_Caffe_Different_Width_Height)
#if defined(INF_ENGINE_RELEASE)
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 || backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) &&
target == DNN_TARGET_MYRIAD && getInferenceEngineVPUType() == CV_DNN_INFERENCE_ENGINE_VPU_TYPE_MYRIAD_X)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD_X);
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD_X, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE exception: Ngraph operation Transpose with name conv15_2_mbox_conf_perm has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
Mat sample = imread(findDataFile("dnn/street.png"));
Mat inp = blobFromImage(sample, 1.0f / 127.5, Size(300, 560), Scalar(127.5, 127.5, 127.5), false);
float scoreDiff = 0.0, iouDiff = 0.0;
@ -324,12 +332,14 @@ TEST_P(DNNTestNetwork, SSD_VGG16)
CV_TEST_TAG_DEBUG_VERYLONG);
if (backend == DNN_BACKEND_HALIDE && target == DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_HALIDE); // TODO HALIDE_CPU
Mat sample = imread(findDataFile("dnn/street.png"));
Mat inp = blobFromImage(sample, 1.0f, Size(300, 300), Scalar(), false);
float scoreDiff = 0.0, iouDiff = 0.0;
if (target == DNN_TARGET_OPENCL_FP16)
{
scoreDiff = 0.0325;
scoreDiff = 0.04;
}
else if (target == DNN_TARGET_MYRIAD)
{

16
modules/dnn/test/test_caffe_importer.cpp
View File

@ -517,10 +517,12 @@ TEST_P(Test_Caffe_nets, Colorization)
l1 = 0.21;
lInf = 4.5;
}
#if defined(INF_ENGINE_RELEASE)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
l1 = 0.26; lInf = 6.5;
l1 = 0.3; lInf = 10;
}
#endif
normAssert(out, ref, "", l1, lInf);
expectNoFallbacksFromIE(net);
@ -713,6 +715,13 @@ TEST_P(Test_Caffe_nets, FasterRCNN_zf)
#endif
CV_TEST_TAG_DEBUG_LONG
);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE exception: Ngraph operation Reshape with name rpn_cls_score_reshape has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 ||
backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16);
@ -734,6 +743,11 @@ TEST_P(Test_Caffe_nets, RFCN)
CV_TEST_TAG_LONG,
CV_TEST_TAG_DEBUG_VERYLONG
);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// Exception: Function contains several inputs and outputs with one friendly name! (HETERO bug?)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 ||
backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16);

5
modules/dnn/test/test_common.hpp
View File

@ -172,16 +172,19 @@ public:
static void checkBackend(int backend, int target, Mat* inp = 0, Mat* ref = 0)
{
CV_UNUSED(backend); CV_UNUSED(target); CV_UNUSED(inp); CV_UNUSED(ref);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021000000)
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 || backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
&& target == DNN_TARGET_MYRIAD)
{
if (inp && ref && inp->dims == 4 && ref->dims == 4 &&
inp->size[0] != 1 && inp->size[0] != ref->size[0])
{
std::cout << "Inconsistent batch size of input and output blobs for Myriad plugin" << std::endl;
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD);
throw SkipTestException("Inconsistent batch size of input and output blobs for Myriad plugin");
}
}
#endif
}
void expectNoFallbacks(Net& net, bool raiseError = true)

55
modules/dnn/test/test_darknet_importer.cpp
View File

@ -245,13 +245,13 @@ public:
nms_boxes.push_back(box);
nms_confidences.push_back(conf);
nms_classIds.push_back(class_id);
#if 0 // use to update test reference data
std::cout << b << ", " << class_id << ", " << conf << "f, "
<< box.x << "f, " << box.y << "f, "
<< box.x + box.width << "f, " << box.y + box.height << "f,"
<< std::endl;
#endif
if (cvtest::debugLevel > 0)
{
std::cout << b << ", " << class_id << ", " << conf << "f, "
<< box.x << "f, " << box.y << "f, "
<< box.x + box.width << "f, " << box.y + box.height << "f,"
<< std::endl;
}
}
if (cvIsNaN(iouDiff))
@ -359,6 +359,13 @@ TEST_P(Test_Darknet_nets, YoloVoc)
scoreDiff = 0.03;
iouDiff = 0.018;
}
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
iouDiff = std::numeric_limits<double>::quiet_NaN();
}
#endif
std::string config_file = "yolo-voc.cfg";
std::string weights_file = "yolo-voc.weights";
@ -372,6 +379,12 @@ TEST_P(Test_Darknet_nets, YoloVoc)
SCOPED_TRACE("batch size 2");
testDarknetModel(config_file, weights_file, ref, scoreDiff, iouDiff, 0.24, nmsThreshold);
}
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
}
TEST_P(Test_Darknet_nets, TinyYoloVoc)
@ -615,6 +628,14 @@ TEST_P(Test_Darknet_nets, YOLOv4)
std::string config_file = "yolov4.cfg";
std::string weights_file = "yolov4.weights";
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy (batch 1)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
iouDiff = std::numeric_limits<double>::quiet_NaN();
}
#endif
#if defined(INF_ENGINE_RELEASE)
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 ||
backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) && target == DNN_TARGET_MYRIAD &&
@ -633,6 +654,13 @@ TEST_P(Test_Darknet_nets, YOLOv4)
{
SCOPED_TRACE("batch size 2");
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy (batch 1)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
iouDiff = 0.45f;
}
#endif
#if defined(INF_ENGINE_RELEASE)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
@ -648,6 +676,12 @@ TEST_P(Test_Darknet_nets, YOLOv4)
testDarknetModel(config_file, weights_file, ref, scoreDiff, iouDiff);
}
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
}
TEST_P(Test_Darknet_nets, YOLOv4_tiny)
@ -718,6 +752,13 @@ TEST_P(Test_Darknet_nets, YOLOv4x_mish)
{
applyTestTag(CV_TEST_TAG_LONG, (target == DNN_TARGET_CPU ? CV_TEST_TAG_MEMORY_1GB : CV_TEST_TAG_MEMORY_2GB));
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE exception: Ngraph operation Transpose with name permute_168 has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2020040000) // nGraph compilation failure
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_VERSION);

21
modules/dnn/test/test_halide_layers.cpp
View File

@ -39,12 +39,13 @@ static void test(Mat& input, Net& net, Backend backendId, Target targetId, bool
l1 = default_l1;
if (lInf == 0.0)
lInf = default_lInf;
#if 0
std::cout << "l1=" << l1 << " lInf=" << lInf << std::endl;
std::cout << outputDefault.reshape(1, outputDefault.total()).t() << std::endl;
std::cout << outputHalide.reshape(1, outputDefault.total()).t() << std::endl;
#endif
normAssert(outputDefault, outputHalide, "", l1, lInf);
if (cvtest::debugLevel > 0 || testing::Test::HasFailure())
{
std::cout << "l1=" << l1 << " lInf=" << lInf << std::endl;
std::cout << outputDefault.reshape(1, outputDefault.total()).t() << std::endl;
std::cout << outputHalide.reshape(1, outputDefault.total()).t() << std::endl;
}
}
static void test(LayerParams& params, Mat& input, Backend backendId, Target targetId, bool skipCheck = false, double l1 = 0.0, double lInf = 0.0)
@ -802,6 +803,16 @@ TEST_P(Eltwise, Accuracy)
Backend backendId = get<0>(get<4>(GetParam()));
Target targetId = get<1>(get<4>(GetParam()));
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && targetId == DNN_TARGET_OPENCL &&
inSize == Vec3i(1, 4, 5) && op == "sum" && numConv == 1 && !weighted)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && targetId == DNN_TARGET_OPENCL &&
inSize == Vec3i(2, 8, 6) && op == "sum" && numConv == 1 && !weighted)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LE(2018050000)
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && targetId == DNN_TARGET_MYRIAD &&
inSize == Vec3i(1, 4, 5))

10
modules/dnn/test/test_layers.cpp
View File

@ -378,7 +378,7 @@ TEST_P(Test_Caffe_layers, layer_prelu_fc)
// Reference output values are in range [-0.0001, 10.3906]
double l1 = (target == DNN_TARGET_MYRIAD) ? 0.005 : 0.0;
double lInf = (target == DNN_TARGET_MYRIAD) ? 0.021 : 0.0;
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2020040000)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2020040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL)
{
l1 = 0.006f; lInf = 0.05f;
@ -1475,6 +1475,14 @@ TEST_P(Test_DLDT_two_inputs, as_backend)
lInf = 0.3;
}
normAssert(out, ref, "", l1, lInf);
if (cvtest::debugLevel > 0 || HasFailure())
{
std::cout << "input1 scale=" << kScale << " input2 scale=" << kScaleInv << std::endl;
std::cout << "input1: " << firstInp.size << " " << firstInp.reshape(1, 1) << std::endl;
std::cout << "input2: " << secondInp.size << " " << secondInp.reshape(1, 1) << std::endl;
std::cout << "ref: " << ref.reshape(1, 1) << std::endl;
std::cout << "out: " << out.reshape(1, 1) << std::endl;
}
}
INSTANTIATE_TEST_CASE_P(/*nothing*/, Test_DLDT_two_inputs, Combine(

43
modules/dnn/test/test_onnx_importer.cpp
View File

@ -802,6 +802,14 @@ TEST_P(Test_ONNX_layers, Split_EltwiseMax)
TEST_P(Test_ONNX_layers, LSTM_Activations)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE Exception: Ngraph operation Reshape with name Block1237_Output_0_before_reshape has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
testONNXModels("lstm_cntk_tanh", pb, 0, 0, false, false);
}
@ -946,6 +954,13 @@ TEST_P(Test_ONNX_layers, Conv1d_variable_weight_bias)
TEST_P(Test_ONNX_layers, GatherMultiOutput)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE Exception: Ngraph operation Reshape with name 6 has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021030000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // exception
@ -953,7 +968,7 @@ TEST_P(Test_ONNX_layers, GatherMultiOutput)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // exception
#endif
#if defined(INF_ENGINE_RELEASE)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LE(2021030000)
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE);
#endif
@ -963,14 +978,25 @@ TEST_P(Test_ONNX_layers, GatherMultiOutput)
TEST_P(Test_ONNX_layers, DynamicAxes)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
#if defined(INF_ENGINE_RELEASE)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
}
#if INF_ENGINE_VER_MAJOR_LT(2021000000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
}
#endif
#endif
testONNXModels("squeeze_and_conv_dynamic_axes");
testONNXModels("unsqueeze_and_conv_dynamic_axes");
testONNXModels("gather_dynamic_axes");
@ -1050,6 +1076,13 @@ TEST_P(Test_ONNX_layers, PoolConv1d)
TEST_P(Test_ONNX_layers, ConvResizePool1d)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE Exception: Ngraph operation Reshape with name 15 has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
#if defined(INF_ENGINE_RELEASE)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
@ -1367,8 +1400,8 @@ TEST_P(Test_ONNX_nets, TinyYolov2)
double l1 = default_l1, lInf = default_lInf;
if (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD)
{
l1 = 0.017;
lInf = 0.14;
l1 = 0.02;
lInf = 0.2;
}
else if (target == DNN_TARGET_CUDA_FP16)
{
@ -1465,10 +1498,10 @@ TEST_P(Test_ONNX_nets, Emotion_ferplus)
l1 = 2.4e-4;
lInf = 6e-4;
}
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2020040000)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2020040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
l1 = 0.012f; lInf = 0.035f;
l1 = 0.013f; lInf = 0.035f;
}
#endif

256
modules/dnn/test/test_tf_importer.cpp
View File

@ -84,6 +84,10 @@ public:
void runTensorFlowNet(const std::string& prefix, bool hasText = false,
double l1 = 0.0, double lInf = 0.0, bool memoryLoad = false, const std::string& groupPrefix = "")
{
if (cvtest::debugLevel > 0)
{
std::cout << prefix << groupPrefix << std::endl;
}
std::string netPath = path(prefix + groupPrefix + "_net.pb");
std::string netConfig = (hasText ? path(prefix + groupPrefix + "_net.pbtxt") : "");
std::string inpPath = path(prefix + "_in.npy");
@ -119,6 +123,16 @@ public:
net.setInput(input);
cv::Mat output = net.forward();
normAssert(ref, output, "", l1 ? l1 : default_l1, lInf ? lInf : default_lInf);
if (cvtest::debugLevel > 0 || HasFailure())
{
std::cout << "input: " << input.size << std::endl;
std::cout << input.reshape(1, 1) << std::endl;
std::cout << "ref " << ref.size << std::endl;
std::cout << ref.reshape(1, 1) << std::endl;
std::cout << "output: " << output.size << std::endl;
std::cout << output.reshape(1, 1) << std::endl;
}
}
};
@ -133,7 +147,7 @@ TEST_P(Test_TensorFlow_layers, reduce_max)
{
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
runTensorFlowNet("max_pool_by_axis");
runTensorFlowNet("max_pool_by_axis", false, 0.0f, 0.0f);
}
TEST_P(Test_TensorFlow_layers, reduce_sum)
@ -145,7 +159,11 @@ TEST_P(Test_TensorFlow_layers, reduce_sum)
TEST_P(Test_TensorFlow_layers, reduce_max_channel)
{
runTensorFlowNet("reduce_max_channel");
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2020040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD) // incorrect result
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
runTensorFlowNet("reduce_max_channel", false, 0.0f, 0.0f);
}
TEST_P(Test_TensorFlow_layers, reduce_sum_channel)
@ -155,6 +173,10 @@ TEST_P(Test_TensorFlow_layers, reduce_sum_channel)
TEST_P(Test_TensorFlow_layers, reduce_max_channel_keep_dims)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2020040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD) // incorrect result
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
runTensorFlowNet("reduce_max_channel", false, 0.0, 0.0, false, "_keep_dims");
}
@ -221,13 +243,49 @@ TEST_P(Test_TensorFlow_layers, padding)
runTensorFlowNet("keras_pad_concat");
}
TEST_P(Test_TensorFlow_layers, padding_asymmetric)
TEST_P(Test_TensorFlow_layers, padding_asymmetric_1)
{
runTensorFlowNet("conv2d_asymmetric_pads_nchw");
}
TEST_P(Test_TensorFlow_layers, padding_asymmetric_2)
{
runTensorFlowNet("conv2d_asymmetric_pads_nhwc");
}
TEST_P(Test_TensorFlow_layers, padding_asymmetric_3)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_CPU) // Exception: Unsupported pad value
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_CPU, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2020020000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD) // Exception: Unsupported pad value
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
runTensorFlowNet("max_pool2d_asymmetric_pads_nchw");
}
TEST_P(Test_TensorFlow_layers, padding_asymmetric_4)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_CPU) // Exception: Unsupported pad value
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_CPU, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2020020000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD) // Exception: Unsupported pad value
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
runTensorFlowNet("max_pool2d_asymmetric_pads_nhwc");
}
TEST_P(Test_TensorFlow_layers, padding_asymmetric_5)
{
runTensorFlowNet("conv2d_backprop_input_asymmetric_pads_nchw");
}
TEST_P(Test_TensorFlow_layers, padding_asymmetric_6)
{
runTensorFlowNet("conv2d_backprop_input_asymmetric_pads_nhwc");
}
@ -268,6 +326,13 @@ TEST_P(Test_TensorFlow_layers, pad_and_concat)
TEST_P(Test_TensorFlow_layers, concat_axis_1)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE Exception: Ngraph operation Transpose with name Flatten_1/flatten/Reshape/nhwc has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021030000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // exception
@ -423,19 +488,77 @@ TEST_P(Test_TensorFlow_layers, pooling_reduce_sum)
runTensorFlowNet("reduce_sum"); // a SUM pooling over all spatial dimensions.
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum2)
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_0_false)
{
int axises[] = {0, 1, 2, 3};
for (int keepdims = 0; keepdims <= 1; ++keepdims)
{
for (int i = 0; i < sizeof(axises)/sizeof(axises[0]); ++i)
{
runTensorFlowNet(cv::format("reduce_sum_%d_%s", axises[i], (keepdims ? "True" : "False")));
}
runTensorFlowNet(cv::format("reduce_sum_1_2_%s", keepdims ? "True" : "False"));
}
runTensorFlowNet("reduce_sum_0_False");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_1_false)
{
runTensorFlowNet("reduce_sum_1_False");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_2_false)
{
runTensorFlowNet("reduce_sum_2_False");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_3_false)
{
runTensorFlowNet("reduce_sum_3_False");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_1_2_false)
{
#if defined(INF_ENGINE_RELEASE)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD)
{
default_l1 = 0.01f;
}
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
default_l1 = 0.01f;
}
#endif
runTensorFlowNet("reduce_sum_1_2_False");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_0_true)
{
runTensorFlowNet("reduce_sum_0_True");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_1_true)
{
runTensorFlowNet("reduce_sum_1_True");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_2_true)
{
runTensorFlowNet("reduce_sum_2_True");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_3_true)
{
runTensorFlowNet("reduce_sum_3_True");
}
TEST_P(Test_TensorFlow_layers, pooling_reduce_sum_1_2_true)
{
#if defined(INF_ENGINE_RELEASE)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD)
{
default_l1 = 0.01f;
}
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
default_l1 = 0.01f;
}
#endif
runTensorFlowNet("reduce_sum_1_2_True");
}
TEST_P(Test_TensorFlow_layers, max_pool_grad)
{
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
@ -715,13 +838,14 @@ TEST_P(Test_TensorFlow_nets, MobileNet_SSD)
double scoreDiff = default_l1, iouDiff = default_lInf;
if (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD)
{
scoreDiff = 0.0043;
iouDiff = 0.037;
scoreDiff = 0.01;
iouDiff = 0.1;
}
else if (target == DNN_TARGET_CUDA_FP16)
{
iouDiff = 0.04;
}
normAssertDetections(ref, out, "", 0.2, scoreDiff, iouDiff);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_RELEASE >= 2019010000
expectNoFallbacksFromIE(net);
@ -815,16 +939,13 @@ TEST_P(Test_TensorFlow_nets, MobileNet_v1_SSD)
expectNoFallbacksFromIE(net);
}
TEST_P(Test_TensorFlow_nets, Faster_RCNN)
TEST_P(Test_TensorFlow_nets, Faster_RCNN_inception_v2_coco_2018_01_28)
{
// FIXIT split test
applyTestTag(
(target == DNN_TARGET_CPU ? CV_TEST_TAG_MEMORY_1GB : CV_TEST_TAG_MEMORY_2GB),
CV_TEST_TAG_LONG,
CV_TEST_TAG_DEBUG_VERYLONG
);
static std::string names[] = {"faster_rcnn_inception_v2_coco_2018_01_28",
"faster_rcnn_resnet50_coco_2018_01_28"};
#ifdef INF_ENGINE_RELEASE
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 &&
@ -835,13 +956,82 @@ TEST_P(Test_TensorFlow_nets, Faster_RCNN)
backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
// segfault: inference-engine/thirdparty/clDNN/src/gpu/detection_output_cpu.cpp:111:
// Assertion `prior_height > 0' failed.
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
if (backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL_FP16);
if (backend == DNN_BACKEND_CUDA && target == DNN_TARGET_CUDA_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_CUDA_FP16);
checkBackend();
double scoresDiff = 1e-5;
double iouDiff = 1e-4;
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 || backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
scoresDiff = 0.02;
iouDiff = 0.1;
}
std::string name = "faster_rcnn_inception_v2_coco_2018_01_28";
{
std::string proto = findDataFile("dnn/" + name + ".pbtxt");
std::string model = findDataFile("dnn/" + name + ".pb", false);
Net net = readNetFromTensorflow(model, proto);
net.setPreferableBackend(backend);
net.setPreferableTarget(target);
Mat img = imread(findDataFile("dnn/dog416.png"));
Mat blob = blobFromImage(img, 1.0f, Size(800, 600), Scalar(), true, false);
net.setInput(blob);
Mat out = net.forward();
Mat ref = blobFromNPY(findDataFile("dnn/tensorflow/" + name + ".detection_out.npy"));
// accuracy (both OpenCV & IE)
if (target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL_FP16);
normAssertDetections(ref, out, name.c_str(), 0.3, scoresDiff, iouDiff);
}
}
TEST_P(Test_TensorFlow_nets, Faster_RCNN_resnet50_coco_2018_01_28)
{
applyTestTag(
(target == DNN_TARGET_CPU ? CV_TEST_TAG_MEMORY_1GB : CV_TEST_TAG_MEMORY_2GB),
CV_TEST_TAG_LONG,
CV_TEST_TAG_DEBUG_VERYLONG
);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE exception: Ngraph operation Transpose with name FirstStageBoxPredictor/ClassPredictor/reshape_1/nhwc has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
#ifdef INF_ENGINE_RELEASE
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 &&
(INF_ENGINE_VER_MAJOR_LT(2019020000) || target != DNN_TARGET_CPU))
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
if (INF_ENGINE_VER_MAJOR_GT(2019030000) &&
backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
// segfault: inference-engine/thirdparty/clDNN/src/gpu/detection_output_cpu.cpp:111:
// Assertion `prior_height > 0' failed.
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
if (backend == DNN_BACKEND_CUDA && target == DNN_TARGET_CUDA_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_CUDA_FP16);
@ -856,10 +1046,11 @@ TEST_P(Test_TensorFlow_nets, Faster_RCNN)
scoresDiff = 0.06;
iouDiff = 0.08;
}
for (int i = 0; i < 2; ++i)
std::string name = "faster_rcnn_resnet50_coco_2018_01_28";
{
std::string proto = findDataFile("dnn/" + names[i] + ".pbtxt");
std::string model = findDataFile("dnn/" + names[i] + ".pb", false);
std::string proto = findDataFile("dnn/" + name + ".pbtxt");
std::string model = findDataFile("dnn/" + name + ".pb", false);
Net net = readNetFromTensorflow(model, proto);
net.setPreferableBackend(backend);
@ -870,8 +1061,13 @@ TEST_P(Test_TensorFlow_nets, Faster_RCNN)
net.setInput(blob);
Mat out = net.forward();
Mat ref = blobFromNPY(findDataFile("dnn/tensorflow/" + names[i] + ".detection_out.npy"));
normAssertDetections(ref, out, names[i].c_str(), 0.3, scoresDiff, iouDiff);
Mat ref = blobFromNPY(findDataFile("dnn/tensorflow/" + name + ".detection_out.npy"));
// accuracy
if (target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL_FP16);
normAssertDetections(ref, out, name.c_str(), 0.3, scoresDiff, iouDiff);
}
}
@ -1282,6 +1478,10 @@ TEST_P(Test_TensorFlow_layers, resize_bilinear_down)
TEST_P(Test_TensorFlow_layers, resize_concat_optimization)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU) // Exception: Function contains several inputs and outputs with one friendly name! (HETERO bug?)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
runTensorFlowNet("resize_concat_optimization");
}
@ -1406,7 +1606,7 @@ TEST_P(Test_TensorFlow_nets, Mask_RCNN)
Mat outDetections = outs[0];
Mat outMasks = outs[1];
double scoreDiff = (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD) ? 0.019 : 2e-5;
double scoreDiff = (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD) ? 0.2 : 2e-5;
double iouDiff = (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD) ? 0.018 : default_lInf;
normAssertDetections(refDetections, outDetections, "", /*threshold for zero confidence*/1e-5, scoreDiff, iouDiff);
@ -1440,7 +1640,7 @@ TEST_P(Test_TensorFlow_nets, Mask_RCNN)
double inter = cv::countNonZero(masks & refMasks);
double area = cv::countNonZero(masks | refMasks);
EXPECT_GE(inter / area, 0.99);
EXPECT_GE(inter / area, (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD) ? 0.98 : 0.99);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
expectNoFallbacks(net);

2
modules/dnn/test/test_torch_importer.cpp
View File

@ -186,7 +186,7 @@ TEST_P(Test_Torch_layers, run_concat)
TEST_P(Test_Torch_layers, run_depth_concat)
{
double lInf = 0.0;
if (target == DNN_TARGET_OPENCL_FP16)
if (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD)
{
lInf = 0.032;
}

2
modules/highgui/CMakeLists.txt
View File

@ -66,7 +66,7 @@ if(HAVE_QT)
QT5_ADD_RESOURCES(_RCC_OUTFILES ${CMAKE_CURRENT_LIST_DIR}/src/window_QT.qrc)
QT5_WRAP_CPP(_MOC_OUTFILES ${CMAKE_CURRENT_LIST_DIR}/src/window_QT.h)
else()
message(FATAL_ERROR "Unsuported QT version: ${QT_VERSION_MAJOR}")
message(FATAL_ERROR "Unsupported QT version: ${QT_VERSION_MAJOR}")
endif()
list(APPEND highgui_srcs

4
modules/js/generator/embindgen.py
View File

@ -485,7 +485,7 @@ class JSWrapperGenerator(object):
arg_types.append(arg_type)
unwrapped_arg_types.append(arg_type)
# Function attribure
# Function attribute
func_attribs = ''
if '*' in ''.join(arg_types):
func_attribs += ', allow_raw_pointers()'
@ -680,7 +680,7 @@ class JSWrapperGenerator(object):
def_args.append(arg.defval)
arg_types.append(orig_arg_types[-1])
# Function attribure
# Function attribute
func_attribs = ''
if '*' in ''.join(orig_arg_types):
func_attribs += ', allow_raw_pointers()'

17
modules/ts/misc/chart.py
View File

@ -1,5 +1,6 @@
#!/usr/bin/env python
from __future__ import print_function
import testlog_parser, sys, os, xml, re
from table_formatter import *
from optparse import OptionParser
@ -116,7 +117,7 @@ if __name__ == "__main__":
(options, args) = parser.parse_args()
if len(args) != 1:
print >> sys.stderr, "Usage:\n", os.path.basename(sys.argv[0]), "<log_name1>.xml"
print("Usage:\n", os.path.basename(sys.argv[0]), "<log_name1>.xml", file=sys.stderr)
exit(1)
options.generateHtml = detectHtmlOutputType(options.format)
@ -136,7 +137,7 @@ if __name__ == "__main__":
args[0] = os.path.basename(args[0])
if not tests:
print >> sys.stderr, "Error - no tests matched"
print("Error - no tests matched", file=sys.stderr)
exit(1)
argsnum = len(tests[0][1])
@ -156,26 +157,26 @@ if __name__ == "__main__":
names1.add(sn)
if sn == sname:
if len(pair[1]) != argsnum:
print >> sys.stderr, "Error - unable to create chart tables for functions having different argument numbers"
print("Error - unable to create chart tables for functions having different argument numbers", file=sys.stderr)
sys.exit(1)
for i in range(argsnum):
arglists[i][pair[1][i]] = 1
if names1 or len(names) != 1:
print >> sys.stderr, "Error - unable to create tables for functions from different test suits:"
print("Error - unable to create tables for functions from different test suits:", file=sys.stderr)
i = 1
for name in sorted(names):
print >> sys.stderr, "%4s: %s" % (i, name)
print("%4s: %s" % (i, name), file=sys.stderr)
i += 1
if names1:
print >> sys.stderr, "Other suits in this log (can not be chosen):"
print("Other suits in this log (can not be chosen):", file=sys.stderr)
for name in sorted(names1):
print >> sys.stderr, "%4s: %s" % (i, name)
print("%4s: %s" % (i, name), file=sys.stderr)
i += 1
sys.exit(1)
if argsnum < 2:
print >> sys.stderr, "Error - tests from %s have less than 2 parameters" % sname
print("Error - tests from %s have less than 2 parameters" % sname, file=sys.stderr)
exit(1)
for i in range(argsnum):

3
modules/ts/misc/report.py
View File

@ -1,5 +1,6 @@
#!/usr/bin/env python
from __future__ import print_function
import testlog_parser, sys, os, xml, re, glob
from table_formatter import *
from optparse import OptionParser
@ -14,7 +15,7 @@ if __name__ == "__main__":
(options, args) = parser.parse_args()
if len(args) < 1:
print >> sys.stderr, "Usage:\n", os.path.basename(sys.argv[0]), "<log_name1>.xml"
print("Usage:\n", os.path.basename(sys.argv[0]), "<log_name1>.xml", file=sys.stderr)
exit(0)
options.generateHtml = detectHtmlOutputType(options.format)

2
modules/ts/misc/run.py
View File

@ -105,7 +105,7 @@ if __name__ == "__main__":
path = args.build_path
try:
if not os.path.isdir(path):
raise Err("Not a directory (should contain CMakeCache.txt ot test executables)")
raise Err("Not a directory (should contain CMakeCache.txt to test executables)")
cache = CMakeCache(args.configuration)
fname = os.path.join(path, "CMakeCache.txt")

3
modules/ts/misc/summary.py
View File

@ -1,5 +1,6 @@
#!/usr/bin/env python
from __future__ import print_function
import testlog_parser, sys, os, xml, glob, re
from table_formatter import *
from optparse import OptionParser
@ -26,7 +27,7 @@ def getSetName(tset, idx, columns, short = True):
if __name__ == "__main__":
if len(sys.argv) < 2:
print >> sys.stderr, "Usage:\n", os.path.basename(sys.argv[0]), "<log_name1>.xml [<log_name2>.xml ...]"
print("Usage:\n", os.path.basename(sys.argv[0]), "<log_name1>.xml [<log_name2>.xml ...]", file=sys.stderr)
exit(0)
parser = OptionParser()

4
platforms/winpack_dldt/build_package.py
View File

@ -163,7 +163,7 @@ class BuilderDLDT:
self.config = config
cpath = self.config.dldt_config
log.info('DLDT build configration: %s', cpath)
log.info('DLDT build configuration: %s', cpath)
if not os.path.exists(cpath):
cpath = os.path.join(SCRIPT_DIR, cpath)
if not os.path.exists(cpath):
@ -575,5 +575,5 @@ if __name__ == "__main__":
try:
main()
except:
log.info('FATAL: Error occured. To investigate problem try to change logging level using LOGLEVEL=DEBUG environment variable.')
log.info('FATAL: Error occurred. To investigate problem try to change logging level using LOGLEVEL=DEBUG environment variable.')
raise

2
samples/dnn/siamrpnpp.py
View File

@ -300,7 +300,7 @@ class SiamRPNTracker:
# clip boundary
cx, cy, width, height = self._bbox_clip(cx, cy, width, height, img.shape[:2])
# udpate state
# update state
self.center_pos = np.array([cx, cy])
self.w = width
self.h = height

6
samples/dnn/tf_text_graph_ssd.py
View File

@ -270,12 +270,12 @@ def createSSDGraph(modelPath, configPath, outputPath):
addConstNode('concat/axis_flatten', [-1], graph_def)
addConstNode('PriorBox/concat/axis', [-2], graph_def)
for label in ['ClassPredictor', 'BoxEncodingPredictor' if box_predictor is 'convolutional' else 'BoxPredictor']:
for label in ['ClassPredictor', 'BoxEncodingPredictor' if box_predictor == 'convolutional' else 'BoxPredictor']:
concatInputs = []
for i in range(num_layers):
# Flatten predictions
flatten = NodeDef()
if box_predictor is 'convolutional':
if box_predictor == 'convolutional':
inpName = 'BoxPredictor_%d/%s/BiasAdd' % (i, label)
else:
if i == 0:
@ -308,7 +308,7 @@ def createSSDGraph(modelPath, configPath, outputPath):
priorBox = NodeDef()
priorBox.name = 'PriorBox_%d' % i
priorBox.op = 'PriorBox'
if box_predictor is 'convolutional':
if box_predictor == 'convolutional':
priorBox.input.append('BoxPredictor_%d/BoxEncodingPredictor/BiasAdd' % i)
else:
if i == 0:

2
samples/python/tutorial_code/video/meanshift/camshift.py
View File

@ -24,7 +24,7 @@ mask = cv.inRange(hsv_roi, np.array((0., 60.,32.)), np.array((180.,255.,255.)))
roi_hist = cv.calcHist([hsv_roi],[0],mask,[180],[0,180])
cv.normalize(roi_hist,roi_hist,0,255,cv.NORM_MINMAX)
# Setup the termination criteria, either 10 iteration or move by atleast 1 pt
# Setup the termination criteria, either 10 iteration or move by at least 1 pt
term_crit = ( cv.TERM_CRITERIA_EPS | cv.TERM_CRITERIA_COUNT, 10, 1 )
while(1):

2
samples/python/tutorial_code/video/meanshift/meanshift.py
View File

@ -24,7 +24,7 @@ mask = cv.inRange(hsv_roi, np.array((0., 60.,32.)), np.array((180.,255.,255.)))
roi_hist = cv.calcHist([hsv_roi],[0],mask,[180],[0,180])
cv.normalize(roi_hist,roi_hist,0,255,cv.NORM_MINMAX)
# Setup the termination criteria, either 10 iteration or move by atleast 1 pt
# Setup the termination criteria, either 10 iteration or move by at least 1 pt
term_crit = ( cv.TERM_CRITERIA_EPS | cv.TERM_CRITERIA_COUNT, 10, 1 )
while(1):