Open Source Computer Vision Library
Go to file
Daniele Affinita 2a333a6c86
Merge pull request #25644 from DaniAffCH:blockwise-quantization
[GSoC] dnn: Blockwise quantization support #25644

This PR introduces blockwise quantization in DNN allowing the parsing of ONNX models quantized in blockwise style. In particular it modifies the `Quantize` and `Dequantize` operations. The related PR opencv/opencv_extra#1181 contains the test data.

Additional notes:
- The original quantization issue has been fixed. Previously, for 1D scale and zero-point, the operation applied was  $y = int8(x/s - z)$ instead of $y = int8(x/s + z)$. Note that the operation was already correctly implemented when the scale and zero-point were scalars. The previous implementation failed the ONNX test cases, but now all have passed successfully.  [Reference](https://github.com/onnx/onnx/blob/main/docs/Operators.md#QuantizeLinear)
- the function `block_repeat` broadcasts scale and zero-point to the input shape. It repeats all the elements of a given axis n times. This function generalizes the behavior of `repeat` from the core module which is defined just for 2 axis assuming `Mat` has 2 dimensions. If appropriate and useful, you might consider moving `block_repeat` to the core module.
- Now, the scale and zero-point can be taken as layer inputs. This increases the ONNX layers' coverage and enables us to run the ONNX test cases (previously disabled) being fully compliant with ONNX standards. Since they are now supported, I have enabled the test cases for: `test_dequantizelinear`, `test_dequantizelinear_axis`, `test_dequantizelinear_blocked`, `test_quantizelinear`, `test_quantizelinear_axis`, `test_quantizelinear_blocked` just in CPU backend. All of them pass successfully.
   
### 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
- [ ] There is a reference to the original bug report and related work
- [x] 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
2024-07-30 14:16:08 +03:00
.github Force contributors to define Apache 2.0 license for the new PRs. 2024-06-28 16:54:14 +03:00
3rdparty Update IPP integration 2024-07-18 08:16:19 -07:00
apps Merge pull request #25582 from fengyuentau:dnn/dump_pbtxt 2024-05-17 11:07:05 +03:00
cmake Merge pull request #25883 from hanliutong:rvv-intrin-upgrade 2024-07-19 11:41:42 +03:00
data Merge pull request #22727 from su77ungr:patch-1 2022-11-17 06:54:25 +00:00
doc Merge pull request #25907 from Andy-SAKN:Andy-SAKN-add_orbbec 2024-07-18 17:27:00 +03:00
include exclude opencv_contrib modules 2020-02-26 15:12:45 +03:00
modules Merge pull request #25644 from DaniAffCH:blockwise-quantization 2024-07-30 14:16:08 +03:00
platforms Fix toolchain. 2024-07-18 13:47:53 +00:00
samples Merge pull request #25822 from mqcmd196:gtk3-gl-support 2024-07-15 17:06:30 +03:00
.editorconfig add .editorconfig 2018-10-11 17:57:51 +00:00
.gitattributes cmake: generate and install ffmpeg-download.ps1 2018-06-09 13:19:48 +03:00
.gitignore Merge pull request #17165 from komakai:objc-binding 2020-06-08 18:32:53 +00:00
CMakeLists.txt cuda: prevent CMAKE_CUDA_ARCHITECTURES default from overiding OpenCV architecture search 2024-07-22 18:00:36 +03:00
CONTRIBUTING.md migration: github.com/opencv/opencv 2016-07-12 12:51:12 +03:00
COPYRIGHT copyright: 2023 (update) 2023-01-09 09:49:22 +00:00
LICENSE Merge pull request #18073 from vpisarev:apache2_license 2020-08-17 11:49:11 +00:00
README.md Space mistake in README.md 2024-03-03 23:37:07 +05:30
SECURITY.md Updated PGP key for security reports 2023-04-19 19:16:55 +03:00

OpenCV: Open Source Computer Vision Library

Resources

Contributing

Please read the contribution guidelines before starting work on a pull request.

Summary of the guidelines:

  • One pull request per issue;
  • Choose the right base branch;
  • Include tests and documentation;
  • Clean up "oops" commits before submitting;
  • Follow the coding style guide.

Additional Resources