potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-31 08:13:04 +04:00
Code Packages Releases Wiki Activity

Merge pull request #24980 from fengyuentau:on-fly-quantization-removal

Browse Source 
dnn cleanup: On-fly-quantization removal #2498

On-fly-quantization is first introduced via https://github.com/opencv/opencv/pull/20228.
We decided to remove it but keep int8 layers implementation because on-fly-quantization
is less practical given the fact that there has been so many dedicated tools for model
quantization.

### 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
- [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
This commit is contained in:
Yuantao Feng
2024-02-16 23:21:45 +08:00
committed by GitHub
parent 3125f9708d
commit d4fd5157fa
27 changed files with 6 additions and 728 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/include/opencv2/dnn/dnn.hpp
-30
View File
@@ -259,15 +259,6 @@ CV__DNN_INLINE_NS_BEGIN
*/
virtual void forward(InputArrayOfArrays inputs, OutputArrayOfArrays outputs, OutputArrayOfArrays internals);
/** @brief Tries to quantize the given layer and compute the quantization parameters required for fixed point implementation.
* @param[in] scales input and output scales.
* @param[in] zeropoints input and output zeropoints.
* @param[out] params Quantized parameters required for fixed point implementation of that layer.
* @returns True if layer can be quantized.
*/
virtual bool tryQuantize(const std::vector<std::vector<float> > &scales,
const std::vector<std::vector<int> > &zeropoints, LayerParams& params);
/** @brief Given the @p input blobs, computes the output @p blobs.
* @param[in] inputs the input blobs.
* @param[out] outputs allocated output blobs, which will store results of the computation.
@@ -610,27 +601,6 @@ CV__DNN_INLINE_NS_BEGIN
CV_WRAP_AS(forwardAndRetrieve) void forward(CV_OUT std::vector<std::vector<Mat> >& outputBlobs,
const std::vector<String>& outBlobNames);
/** @brief Returns a quantized Net from a floating-point Net.
* @param calibData Calibration data to compute the quantization parameters.
* @param inputsDtype Datatype of quantized net's inputs. Can be CV_32F or CV_8S.
* @param outputsDtype Datatype of quantized net's outputs. Can be CV_32F or CV_8S.
* @param perChannel Quantization granularity of quantized Net. The default is true, that means quantize model
* in per-channel way (channel-wise). Set it false to quantize model in per-tensor way (or tensor-wise).
*/
CV_WRAP Net quantize(InputArrayOfArrays calibData, int inputsDtype, int outputsDtype, bool perChannel=true);
/** @brief Returns input scale and zeropoint for a quantized Net.
* @param scales output parameter for returning input scales.
* @param zeropoints output parameter for returning input zeropoints.
*/
CV_WRAP void getInputDetails(CV_OUT std::vector<float>& scales, CV_OUT std::vector<int>& zeropoints) const;
/** @brief Returns output scale and zeropoint for a quantized Net.
* @param scales output parameter for returning output scales.
* @param zeropoints output parameter for returning output zeropoints.
*/
CV_WRAP void getOutputDetails(CV_OUT std::vector<float>& scales, CV_OUT std::vector<int>& zeropoints) const;
/**
* @brief Ask network to use specific computation backend where it supported.
* @param[in] backendId backend identifier.