instruction
stringclasses
9 values
input
stringlengths
279
5.47k
output
stringlengths
6
9.58k
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Review Line:** 71 **Code Changes:** ```diff + MatShape output_shape; + output_shape.reserve(q - 1 + r - batch_dims - last_indices_dim); + for (int i = 0; i < q - 1; ++i) + output_shape.push_back(indices[i]); + for (int i = batch_dims + last_indices_dim; i < r; ++i) + output_shape.push_back(data[i]); + + outputs.assign(1, output_shape); + return false; + } + ```
makes sense to use `output_shape.reserve()` before the loops.
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
please add default branch with assert. It might be useful for debugging,
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Review Line:** 126 **Code Changes:** ```diff + } + + template <typename iT, typename dT> + void forward_impl(const Mat& data, const Mat& indices, Mat& out) + { + CV_Assert(out.isContinuous()); + CV_Assert(indices.isContinuous()); + CV_Assert(data.isContinuous()); + + + const iT* indices_ptr = indices.ptr<iT>(); ```
check also indices and data
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
you do not need vector here. just: ``` int* sliced_indices = indices_ptr + i * last_indices_dim; ```
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
`break` is missing: ``` 2024-09-03T16:01:13.0386984Z /home/ci/opencv/modules/dnn/src/layers/gatherND.cpp:95:17: warning: this statement may fall through [-Wimplicit-fallthrough=] 2024-09-03T16:01:13.0388547Z 95 | switch (dtype) { 2024-09-03T16:01:13.0389213Z | ^~~~~~ 2024-09-03T16:01:13.0390112Z /home/ci/opencv/modules/dnn/src/layers/gatherND.cpp:105:13: note: here 2024-09-03T16:01:13.0391087Z 105 | case CV_64S: 2024-09-03T16:01:13.0391668Z | ^~~~ ```
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
``` /home/ci/opencv/modules/dnn/src/layers/gatherND.cpp:99:56: error: 'float16_t' was not declared in this scope; did you mean 'float_t'? 2024-09-03T15:59:18.3161177Z 99 | case CV_16F: forward_impl<int32_t, float16_t>(data, indices, out); break; ```
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
I propose to use s16 branch for `CV_16F` case to be hardware independent.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/onnx/onnx_importer.cpp **Change Type:** modified **Context:** PR #26106: Support for GatherND layer **Review Line:** 3217 **Code Changes:** ```diff // URL: https://github.com/microsoft/onnxruntime/blob/master/docs/ContribOperators.md @@ -3213,6 +3214,15 @@ void ONNXImporter::parseHardmax(LayerParams& layerParams, const opencv_onnx::Nod addLayer(layerParams, node_proto); } +void ONNXImporter::parseGatherND(LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto) +{ + CV_Assert(node_proto.input_size() == 2); + layerParams.type = "GatherND"; + int batch_dims = layerParams.get<int>("batch_dims", 0); + layerParams.set("batch_dims", batch_dims); ```
Looks like you can simply use `parseSimpleLayers` to parse `GatherND` since it does not take extra operations.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
int16_t is better here to ensure size.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/onnx/onnx_importer.cpp **Change Type:** modified **Context:** PR #26106: Support for GatherND layer **Review Line:** 3217 **Code Changes:** ```diff // URL: https://github.com/microsoft/onnxruntime/blob/master/docs/ContribOperators.md @@ -3213,6 +3214,15 @@ void ONNXImporter::parseHardmax(LayerParams& layerParams, const opencv_onnx::Nod addLayer(layerParams, node_proto); } +void ONNXImporter::parseGatherND(LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto) +{ + CV_Assert(node_proto.input_size() == 2); + layerParams.type = "GatherND"; + int batch_dims = layerParams.get<int>("batch_dims", 0); + layerParams.set("batch_dims", batch_dims); ```
I propose to leave this as it is, for new parser it makes sence
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
This leads to crush during build time
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/gatherND.cpp **Change Type:** added **Context:** PR #26106: Support for GatherND layer **Code Changes:** ```diff @@ -0,0 +1,185 @@ +#include "../precomp.hpp" +#include <opencv2/dnn/shape_utils.hpp> + +namespace cv { namespace dnn { + +class GatherNDLayerImpl CV_FINAL : public GatherNDLayer +{ +public: + GatherNDLayerImpl(const LayerParams& params) ```
I just pushed fix. `indices_ptr` has template type `iT*` and `sliced_indices` should. It's not always `int`. The solution with vector just hides the problem.
You are an expert OpenCV code reviewer specializing in documentation and code clarity. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/text_detection.cpp **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Code Changes:** ```diff @@ -1,21 +1,28 @@ /* - Text detection model: https://github.com/argman/EAST - Download link: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 + Text detection model (EAST): https://github.com/argman/EAST + Download link for EAST model: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 - Text recognition models can be downloaded directly here: + DB detector model: + https://drive.google.com/uc?export=download&id=17_ABp79PlFt9yPCxSaarVc_DKTmrSGGf ```
Not necessary comment
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/text_detection.cpp **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Code Changes:** ```diff @@ -1,21 +1,28 @@ /* - Text detection model: https://github.com/argman/EAST - Download link: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 + Text detection model (EAST): https://github.com/argman/EAST + Download link for EAST model: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 - Text recognition models can be downloaded directly here: + DB detector model: + https://drive.google.com/uc?export=download&id=17_ABp79PlFt9yPCxSaarVc_DKTmrSGGf ```
The only generic parameter is `Size(width, height)`. Please update `TextDetectionModel_EAST_Impl` constructor with the rest of preprocessing options so in sample you just need to call `detector.setInputSize()` https://github.com/opencv/opencv/blob/f2c3d4dfe3843898d969a45e43d981e8af967bd5/modules/dnn/src/model.cpp#L1115
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/text_detection.cpp **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Code Changes:** ```diff @@ -1,21 +1,28 @@ /* - Text detection model: https://github.com/argman/EAST - Download link: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 + Text detection model (EAST): https://github.com/argman/EAST + Download link for EAST model: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 - Text recognition models can be downloaded directly here: + DB detector model: + https://drive.google.com/uc?export=download&id=17_ABp79PlFt9yPCxSaarVc_DKTmrSGGf ```
Same request as for EAST. Let's move magic constants to `TextDetectionModel_DB` constructor
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/text_detection.cpp **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Code Changes:** ```diff @@ -1,21 +1,28 @@ /* - Text detection model: https://github.com/argman/EAST - Download link: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 + Text detection model (EAST): https://github.com/argman/EAST + Download link for EAST model: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 - Text recognition models can be downloaded directly here: + DB detector model: + https://drive.google.com/uc?export=download&id=17_ABp79PlFt9yPCxSaarVc_DKTmrSGGf ```
Why `frame2` is required?
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/models.yml **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Review Line:** 312 **Code Changes:** ```diff + width: 736 + height: 736 + rgb: false + sample: "text_detection" + +East: + load_info: + url: "https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1" + sha1: "fffabf5ac36f37bddf68e34e84b45f5c4247ed06" + download_name: "frozen_east_text_detection.tar.gz" + download_sha: "3ca8233d6edd748f7ed23246c8ca24cbf696bb94" ```
Exception: Hash mismatch: expected fffabf5ac36f37bddf68e34e84b45f5c4247ed06 vs actual of 3ca8233d6edd748f7ed23246c8ca24cbf696bb94
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/text_detection.cpp **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Code Changes:** ```diff @@ -1,21 +1,28 @@ /* - Text detection model: https://github.com/argman/EAST - Download link: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 + Text detection model (EAST): https://github.com/argman/EAST + Download link for EAST model: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 - Text recognition models can be downloaded directly here: + DB detector model: + https://drive.google.com/uc?export=download&id=17_ABp79PlFt9yPCxSaarVc_DKTmrSGGf ```
There is imageTextR.png, ImageTextN.png with text left.jpg, box_in_scene.png with book covers. They are better candidates for test image.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/text_detection.py **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Code Changes:** ```diff @@ -1,55 +1,82 @@ ''' - Text detection model: https://github.com/argman/EAST - Download link: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 + Text detection model (EAST): https://github.com/argman/EAST + Download link for EAST model: https://www.dropbox.com/s/r2ingd0l3zt8hxs/frozen_east_text_detection.tar.gz?dl=1 - CRNN Text recognition model taken from here: https://github.com/meijieru/crnn.pytorch - How to convert from pb to onnx: - Using classes from here: https://github.com/meijieru/crnn.pytorch/blob/master/models/crnn.py ```
There is no model option in command line params
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/text_detection.py **Change Type:** modified **Context:** PR #25326: Improved and refactored text detection sample in dnn module **Review Line:** 119 **Code Changes:** ```diff + stdImgSize = 512 + imgWidth = min(frame.shape[:2]) + fontSize = (stdSize*imgWidth)/stdImgSize + fontThickness = max(1,(stdWeight*imgWidth)//stdImgSize) + + if(args.alias == "DB"): + # DB Detector initialization + detector = cv2.dnn_TextDetectionModel_DB(args.model) + detector.setBinaryThreshold(args.binary_threshold) + detector.setPolygonThreshold(args.polygon_threshold) + detector.setUnclipRatio(args.unclip_ratio) ```
There is no alias option in command line params
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** cmake/OpenCVCompilerOptimizations.cmake **Change Type:** modified **Context:** PR #26170: 5.x merge 4.x - OpenCV Extra: https://github.com/opencv/opencv_extra/pull/1210 Contrib: https://github.com/opencv/opencv_contrib/pull/3795 ... **Review Line:** 397 **Code Changes:** ```diff ocv_update(CPU_KNOWN_OPTIMIZATIONS "MSA") ocv_update(CPU_MSA_FLAGS_ON "-mmsa") set(CPU_BASELINE "DETECT" CACHE STRING "${HELP_CPU_BASELINE}") + elseif(PPC64LE) + ocv_update(CPU_KNOWN_OPTIMIZATIONS "VSX;VSX3") ocv_update(CPU_VSX_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_vsx.cpp") ocv_update(CPU_VSX3_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_vsx3.cpp") @@ -390,9 +415,6 @@ elseif(PPC64LE) set(CPU_BASELINE "VSX" CACHE STRING "${HELP_CPU_BASELINE}") ```
We need to keep FP16 and RVV_ZVFH in the list of known optimizations (line 55, line 397) as well as TEST_FILE, IMPLIES, FLAGS_ON variables (lines 399-407). I'll adapt flags-riscv64.cmake accordingly after the merge.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** cmake/OpenCVCompilerOptimizations.cmake **Change Type:** modified **Context:** PR #26170: 5.x merge 4.x - OpenCV Extra: https://github.com/opencv/opencv_extra/pull/1210 Contrib: https://github.com/opencv/opencv_contrib/pull/3795 ... **Review Line:** 397 **Code Changes:** ```diff ocv_update(CPU_KNOWN_OPTIMIZATIONS "MSA") ocv_update(CPU_MSA_FLAGS_ON "-mmsa") set(CPU_BASELINE "DETECT" CACHE STRING "${HELP_CPU_BASELINE}") + elseif(PPC64LE) + ocv_update(CPU_KNOWN_OPTIMIZATIONS "VSX;VSX3") ocv_update(CPU_VSX_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_vsx.cpp") ocv_update(CPU_VSX3_TEST_FILE "${OpenCV_SOURCE_DIR}/cmake/checks/cpu_vsx3.cpp") @@ -390,9 +415,6 @@ elseif(PPC64LE) set(CPU_BASELINE "VSX" CACHE STRING "${HELP_CPU_BASELINE}") ```
Lines 55 and 399-407 should also be reverted.
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/cvdef.h **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Review Line:** 972 **Code Changes:** ```diff @@ -968,6 +968,9 @@ CV_INLINE int CV_FOURCC(char c1, char c2, char c3, char c4) return (c1 & 255) + ((c2 & 255) << 8) + ((c3 & 255) << 16) + ((c4 & 255) << 24); } +//! Macro to construct the fourcc code of the codec. Same as CV_FOURCC() +#define CV_FOURCC_MACRO(c1, c2, c3, c4) (((c1) & 255) + (((c2) & 255) << 8) + (((c3) & 255) << 16) + (((c4) & 255) << 24)) + //! @} #ifndef __cplusplus ```
May be move it to videoio.hpp as inline function (which could be exported to bindings) BTW, there is VideoWriter::fourcc (it is confusing because it is used outside of that). So global function is preferable.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/videoio/src/container_avi.cpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Code Changes:** ```diff @@ -2,8 +2,8 @@ // It is subject to the license terms in the LICENSE file found in the top-level directory // of this distribution and at http://opencv.org/license.html. +#include "precomp.hpp" #include "opencv2/videoio/container_avi.private.hpp" -#include <opencv2/core/utils/logger.hpp> #include <fstream> #include <limits> #include <typeinfo> ```
lets put `precomp.hpp` first
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/cvdef.h **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Review Line:** 972 **Code Changes:** ```diff @@ -968,6 +968,9 @@ CV_INLINE int CV_FOURCC(char c1, char c2, char c3, char c4) return (c1 & 255) + ((c2 & 255) << 8) + ((c3 & 255) << 16) + ((c4 & 255) << 24); } +//! Macro to construct the fourcc code of the codec. Same as CV_FOURCC() +#define CV_FOURCC_MACRO(c1, c2, c3, c4) (((c1) & 255) + (((c2) & 255) << 8) + (((c3) & 255) << 16) + (((c4) & 255) << 24)) + //! @} #ifndef __cplusplus ```
There is `CV_FOURCC` inline function above (used in ~70 places). Do you mean move it too? It is used in _imgproc_ for font rendering params. Perhaps we can remove `CV_FOURCC_MACRO` completely, I didn't do it to reduce total changes, because it is used in ~50 places. `VideoWriter::fourcc` is used in ~50 places too.
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/cvdef.h **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Review Line:** 972 **Code Changes:** ```diff @@ -968,6 +968,9 @@ CV_INLINE int CV_FOURCC(char c1, char c2, char c3, char c4) return (c1 & 255) + ((c2 & 255) << 8) + ((c3 & 255) << 16) + ((c4 & 255) << 24); } +//! Macro to construct the fourcc code of the codec. Same as CV_FOURCC() +#define CV_FOURCC_MACRO(c1, c2, c3, c4) (((c1) & 255) + (((c2) & 255) << 8) + (((c3) & 255) << 16) + (((c4) & 255) << 24)) + //! @} #ifndef __cplusplus ```
That is a mess. Lets revise fourcc through a separate task then.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/videoio/src/cap_ximea.cpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Code Changes:** ```diff @@ -8,31 +8,33 @@ #endif using namespace std; +using namespace cv; /**********************************************************************************/ -class CvCaptureCAM_XIMEA : public CvCapture +class CvCaptureCAM_XIMEA : public IVideoCapture ```
.create() could be used instead. Here and several places bellow.
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/videoio/src/precomp.hpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Review Line:** 58 **Code Changes:** ```diff @@ -55,7 +55,6 @@ #include "opencv2/videoio.hpp" -#include "opencv2/videoio/legacy/constants_c.h" #include "opencv2/core/utility.hpp" #ifdef __OPENCV_BUILD @@ -74,7 +73,6 @@ #include "opencv2/imgcodecs.hpp" ```
m.b. removed.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/videoio/misc/gen_dict.json **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Code Changes:** ```diff @@ -1,9 +1,7 @@ { "const_ignore_list": [ "CV_CAP_OPENNI2", - "CV_CAP_PROP_OPENNI_", "CV_CAP_INTELPERC", - "CV_CAP_PROP_INTELPERC_", "CV_CAP_ANY", "CV_CAP_MIL", "CV_CAP_V4L", ```
it should be removed too, right?
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/videoio/src/cap_avfoundation.mm **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Review Line:** 189 **Code Changes:** ```diff int is_color=1); ~CvVideoWriter_AVFoundation(); - bool writeFrame(const IplImage* image) CV_OVERRIDE; + bool isOpened() const CV_OVERRIDE { return mMovieWriter != NULL && mMovieWriter.status != AVAssetWriterStatusFailed; } + void write(cv::InputArray image) CV_OVERRIDE; int getCaptureDomain() const CV_OVERRIDE { return cv::CAP_AVFOUNDATION; } private: - IplImage* argbimage; + cv::Mat argbimage; AVAssetWriter *mMovieWriter; ```
`const cv::Size& `?
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/highgui/src/window_QT.cpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Code Changes:** ```diff @@ -138,12 +138,11 @@ void cv::addText( const Mat& img, const String& text, Point org, const QtFont& f if (!guiMainThread) CV_Error( cv::Error::StsNullPtr, "NULL guiReceiver (please create a window)" ); - CvMat _img = cvMat(img); QMetaObject::invokeMethod( guiMainThread, "putText", autoBlockingConnection(), - Q_ARG(void*, (void*)&_img), ```
`cv::Mat& mat`?
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/highgui/src/window_QT.cpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Code Changes:** ```diff @@ -138,12 +138,11 @@ void cv::addText( const Mat& img, const String& text, Point org, const QtFont& f if (!guiMainThread) CV_Error( cv::Error::StsNullPtr, "NULL guiReceiver (please create a window)" ); - CvMat _img = cvMat(img); QMetaObject::invokeMethod( guiMainThread, "putText", autoBlockingConnection(), - Q_ARG(void*, (void*)&_img), ```
It should release automatically, right?
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/highgui/src/window_QT.cpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Code Changes:** ```diff @@ -138,12 +138,11 @@ void cv::addText( const Mat& img, const String& text, Point org, const QtFont& f if (!guiMainThread) CV_Error( cv::Error::StsNullPtr, "NULL guiReceiver (please create a window)" ); - CvMat _img = cvMat(img); QMetaObject::invokeMethod( guiMainThread, "putText", autoBlockingConnection(), - Q_ARG(void*, (void*)&_img), ```
The same. Release is not needed.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/highgui/src/window_gtk.cpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Review Line:** 359 **Code Changes:** ```diff + image_widget->scaled_image.release(); } - if( !image_widget->scaled_image ){ - image_widget->scaled_image = cvCreateMat( scaled_image_size.height, scaled_image_size.width, CV_8UC3 ); + if( image_widget->scaled_image.empty() ){ + image_widget->scaled_image.create(scaled_image_size.height, scaled_image_size.width, CV_8UC3); } - CV_Assert(image_widget->scaled_image); + CV_Assert(!image_widget->scaled_image.empty()); ```
Release not needed.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/videoio/src/cap_gstreamer.cpp **Change Type:** modified **Context:** PR #26025: C-API cleanup: highgui, videoio - :exclamation: Potential conflicts with #25958 :exclamation: Merge with: opencv/opencv_contrib#3780 This P... **Code Changes:** ```diff @@ -287,10 +287,10 @@ std::string get_gst_propname(int propId) { switch (propId) { - case CV_CAP_PROP_BRIGHTNESS: return "brightness"; - case CV_CAP_PROP_CONTRAST: return "contrast"; - case CV_CAP_PROP_SATURATION: return "saturation"; - case CV_CAP_PROP_HUE: return "hue"; + case CAP_PROP_BRIGHTNESS: return "brightness"; + case CAP_PROP_CONTRAST: return "contrast"; ```
I propose to refresh warning messages here and bellow. `cvWriteFrame` does not exist anymore.
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/perf/opencl/perf_arithm.cpp **Change Type:** modified **Context:** PR #26115: Added more data types to OCL flip() and rotate() perf tests **Review Line:** 361 **Code Changes:** ```diff OCL_PERF_TEST_P(FlipFixture, Flip, ::testing::Combine(OCL_TEST_SIZES, - OCL_TEST_TYPES, FlipType::all())) + ::testing::Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_16UC1, CV_32FC1, CV_32FC4), + FlipType::all())) { const FlipParams params = GetParam(); const Size srcSize = get<0>(params); @@ -387,7 +388,9 @@ typedef tuple<Size, MatType, RotateType> RotateParams; typedef TestBaseWithParam<RotateParams> RotateFixture; ```
You can use `OCL_TEST_TYPES + ::testing::Values(CV_16UC1)` for the both cases.
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/perf/opencl/perf_arithm.cpp **Change Type:** modified **Context:** PR #26115: Added more data types to OCL flip() and rotate() perf tests **Review Line:** 361 **Code Changes:** ```diff OCL_PERF_TEST_P(FlipFixture, Flip, ::testing::Combine(OCL_TEST_SIZES, - OCL_TEST_TYPES, FlipType::all())) + ::testing::Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_16UC1, CV_32FC1, CV_32FC4), + FlipType::all())) { const FlipParams params = GetParam(); const Size srcSize = get<0>(params); @@ -387,7 +388,9 @@ typedef tuple<Size, MatType, RotateType> RotateParams; typedef TestBaseWithParam<RotateParams> RotateFixture; ```
Looks like it does not work in performance tests - implemented for accuracy tests only?
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
cv::Mat::zeros()
You are an expert OpenCV code reviewer specializing in performance optimization. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
Call `parallel_for_` on the outer loop for performance.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/onnx/onnx_importer.cpp **Change Type:** modified **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -197,6 +197,7 @@ class ONNXImporter void parseLayerNorm (LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto); void parseSimpleLayers (LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto); void parseEinsum (LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto); + void parseHardmax (LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto); // Domain: com.microsoft // URL: https://github.com/microsoft/onnxruntime/blob/master/docs/ContribOperators.md @@ -3190,6 +3191,12 @@ void ONNXImporter::parseSimpleLayers(LayerParams& layerParams, const opencv_onnx addLayer(layerParams, node_proto); ```
Dont preprocess axis during importing stage, since we are going to support dynamic shape and shape can be unknown during importing. --- You already called `normalize_axis` in `forward()`. So safe to drop this code.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
FP32 -> FP16 will cause invalid max operation.
You are an expert OpenCV code reviewer specializing in memory management. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
It creates a new memory. You should call `memset` on the pointer of dst to set all zero.
You are an expert OpenCV code reviewer specializing in memory management. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
Returning `true` here stands for input and output can be a shared memory, which is not valid for calling `memset` on dst (because it also sets the input zero). So return `false` here.
You are an expert OpenCV code reviewer specializing in documentation and code clarity. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
Looks like openvino does not have hardmax. See https://docs.openvino.ai/2024/documentation/openvino-ir-format/operation-sets/available-opsets/opset14.html.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
Add `nstripes` behind the lambda function. It equals to total number of loops divided by a constant, e.g. `outer_size * inner_size * (1 / 1024.0)`.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
`src.ptr<const T>`
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
where does `(1/1024.0)` or `1024.0` constant comes from? Just want to understand for myself.
You are an expert OpenCV code reviewer specializing in performance optimization. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + +namespace cv +{ +namespace dnn ```
I learned it from other folks. They said that it is empirical. You could tune it if performance is critical. My understanding is that I want each worker takes care of 1024 times of loops, and there I need `total_loops / 1024` workers in total from the thread pool.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Review Line:** 77 **Code Changes:** ```diff + + axis = normalize_axis(axis, src.dims); + MatShape shape(src.size.p, src.size.p + src.dims); + + // Prepare output + memset(dst.ptr(), 0, dst.total() * dst.elemSize()); + + switch (src.depth()) + { + case CV_8U: hardmaxApply<uchar>(src, dst, axis); break; + case CV_8S: hardmaxApply<schar>(src, dst, axis); break; ```
The change looks redundant. `dst` may be mot contiguous.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Review Line:** 77 **Code Changes:** ```diff + + axis = normalize_axis(axis, src.dims); + MatShape shape(src.size.p, src.size.p + src.dims); + + // Prepare output + memset(dst.ptr(), 0, dst.total() * dst.elemSize()); + + switch (src.depth()) + { + case CV_8U: hardmaxApply<uchar>(src, dst, axis); break; + case CV_8S: hardmaxApply<schar>(src, dst, axis); break; ```
> dst may be mot contiguous. Theoretically it is but I assume it is quite rare in dnn use cases. But anyway, zero can be set inside the loop.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Review Line:** 77 **Code Changes:** ```diff + + axis = normalize_axis(axis, src.dims); + MatShape shape(src.size.p, src.size.p + src.dims); + + // Prepare output + memset(dst.ptr(), 0, dst.total() * dst.elemSize()); + + switch (src.depth()) + { + case CV_8U: hardmaxApply<uchar>(src, dst, axis); break; + case CV_8S: hardmaxApply<schar>(src, dst, axis); break; ```
lets use ::zeros or at least add contigous check.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/harmax.cpp **Change Type:** added **Context:** PR #26079: Add Support for Hardmax Layer **Review Line:** 1 **Code Changes:** ```diff @@ -0,0 +1,140 @@ +#include <inttypes.h> +#include <opencv2/dnn/shape_utils.hpp> +#include "../precomp.hpp" +#include "layers_common.hpp" +#include "../ie_ngraph.hpp" + ```
Just found that filename is wrong, it should be `hardmax_layer.cpp`.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/edge_detection.cpp **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -0,0 +1,242 @@ +#include <opencv2/dnn.hpp> +#include <opencv2/imgproc.hpp> +#include <opencv2/highgui.hpp> +#include <iostream> +#include <string> +#include <cmath> +#include <vector> + +#include "common.hpp" ```
use cv::normalize() with NORM_MINMAX
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/edge_detection.cpp **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -0,0 +1,242 @@ +#include <opencv2/dnn.hpp> +#include <opencv2/imgproc.hpp> +#include <opencv2/highgui.hpp> +#include <iostream> +#include <string> +#include <cmath> +#include <vector> + +#include "common.hpp" ```
remove unnecessary whitespaces
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/edge_detection.cpp **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -0,0 +1,242 @@ +#include <opencv2/dnn.hpp> +#include <opencv2/imgproc.hpp> +#include <opencv2/highgui.hpp> +#include <iostream> +#include <string> +#include <cmath> +#include <vector> + +#include "common.hpp" ```
it would be really nice to be able to switch between Canny and deep learning-based modes with a key
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/edge_detection.cpp **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -0,0 +1,242 @@ +#include <opencv2/dnn.hpp> +#include <opencv2/imgproc.hpp> +#include <opencv2/highgui.hpp> +#include <iostream> +#include <string> +#include <cmath> +#include <vector> + +#include "common.hpp" ```
It should not happen on each iteration, but once before the loop.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/edge_detection.cpp **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -0,0 +1,242 @@ +#include <opencv2/dnn.hpp> +#include <opencv2/imgproc.hpp> +#include <opencv2/highgui.hpp> +#include <iostream> +#include <string> +#include <cmath> +#include <vector> + +#include "common.hpp" ```
Let's use Canny by default, if model is not provided. Also I propose to handle the case in run-time, e.g. show warning, but not fail.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/custom_layer.py **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Review Line:** 1 **Code Changes:** ```diff @@ -0,0 +1,31 @@ +import cv2 as cv + +#! [CropLayer] +class CropLayer(object): + def __init__(self, params, blobs): + self.xstart = 0 ```
The new approach destroys original sample with custom layer. Formally, all code blocks are there, but the sample does not create Net and does not use the new layer.
You are an expert OpenCV code reviewer specializing in API design and compatibility. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/imgproc/include/opencv2/imgproc.hpp **Change Type:** modified **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -1885,7 +1885,7 @@ CV_EXPORTS_W void Laplacian( InputArray src, OutputArray dst, int ddepth, //! @addtogroup imgproc_feature //! @{ -/** @example samples/cpp/edge.cpp +/** @example samples/cpp/snippets/edge.cpp This program demonstrates usage of the Canny edge detector Check @ref tutorial_canny_detector "the corresponding tutorial" for more details ```
We still need simplified API usage example. Code with 200+ lines are just not suitable for initial leaning steps. We need 2 kind of "samples": - simplified API usage example (e.g. move to part of documentation - tutorial_code) - large "task solution" "demos"
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/imgproc/include/opencv2/imgproc.hpp **Change Type:** modified **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -1885,7 +1885,7 @@ CV_EXPORTS_W void Laplacian( InputArray src, OutputArray dst, int ddepth, //! @addtogroup imgproc_feature //! @{ -/** @example samples/cpp/edge.cpp +/** @example samples/cpp/snippets/edge.cpp This program demonstrates usage of the Canny edge detector Check @ref tutorial_canny_detector "the corresponding tutorial" for more details ```
edge.cpp has been moved to samples/cpp/snippets directory as it has been done for other basic cpp samples
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/edge_detection.cpp **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Code Changes:** ```diff @@ -0,0 +1,242 @@ +#include <opencv2/dnn.hpp> +#include <opencv2/imgproc.hpp> +#include <opencv2/highgui.hpp> +#include <iostream> +#include <string> +#include <cmath> +#include <vector> + +#include "common.hpp" ```
fixed, canny is now default
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/edge_detection.cpp **Change Type:** added **Context:** PR #25515: Improved edge detection sample **Review Line:** 13 **Code Changes:** ```diff + +#include "common.hpp" +// Define namespace to simplify code +using namespace cv; +using namespace cv::dnn; +using namespace std; + +int threshold1 = 0; +int threshold2 = 50; +int blurAmount = 5; + ```
Not the best practice in case of `std` namespace
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/imgcodecs/test/test_tiff.cpp **Change Type:** modified **Context:** PR #26093: Update test_tiff.cpp **Review Line:** 1111 **Code Changes:** ```diff @@ -1109,7 +1108,7 @@ TEST(Imgcodecs_Tiff_Modes, write_multipage) vector<Mat> pages; for (size_t i = 0; i < page_count; i++) { - const Mat page = imread(root + page_files[i]); + const Mat page = imread(root + page_files[i], IMREAD_REDUCED_GRAYSCALE_8 + (int)i); pages.push_back(page); } ```
I understood your idea to force output type to grayscale, but why REDUCED?
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/imgcodecs/test/test_tiff.cpp **Change Type:** modified **Context:** PR #26093: Update test_tiff.cpp **Review Line:** 1111 **Code Changes:** ```diff @@ -1109,7 +1108,7 @@ TEST(Imgcodecs_Tiff_Modes, write_multipage) vector<Mat> pages; for (size_t i = 0; i < page_count; i++) { - const Mat page = imread(root + page_files[i]); + const Mat page = imread(root + page_files[i], IMREAD_REDUCED_GRAYSCALE_8 + (int)i); pages.push_back(page); } ```
>I understood your idea to force output type to grayscale, but why REDUCED? to use smaller files to write and speed up the test. ( the code reads 2 gray scale and 4 color images )
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/mat.inl.hpp **Change Type:** modified **Context:** PR #25945: Fix size() for 0d matrix **Review Line:** 1278 **Code Changes:** ```diff @@ -1275,7 +1275,7 @@ Size MatSize::operator()() const { int d = dims(); CV_DbgAssert(d <= 2); - return d == 2 ? Size(p[1], p[0]) : Size(p[0], 1); + return d == 2 ? Size(p[1], p[0]) : d == 1 ? Size(p[0], 1) : Size(0, 0); } inline ```
I believe, scalar should map to `Size(1,1)` rather than `Size(0,0)`. Of course, we should also distinguish between empty mat (the product of all dimensions must be 0) and scalar (the product of all dimensions must be 1).
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/mat.inl.hpp **Change Type:** modified **Context:** PR #25945: Fix size() for 0d matrix **Review Line:** 1278 **Code Changes:** ```diff @@ -1275,7 +1275,7 @@ Size MatSize::operator()() const { int d = dims(); CV_DbgAssert(d <= 2); - return d == 2 ? Size(p[1], p[0]) : Size(p[0], 1); + return d == 2 ? Size(p[1], p[0]) : d == 1 ? Size(p[0], 1) : Size(0, 0); } inline ```
`MatSize` is only used in `cv::Mat` and `cv::UMat` right? What is the scalar you are referring to? Would a matrix of size `1x1x1x1x1x1` be a scalar ? Would size be `(1,1)` then or undefined? An empty matrix can also be `1x0` (`dims()==2`). But when a matrix is newly created, `dims()==0`. It seems weird to also have a size of `(0,1)`
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/mat.inl.hpp **Change Type:** modified **Context:** PR #25945: Fix size() for 0d matrix **Review Line:** 1278 **Code Changes:** ```diff @@ -1275,7 +1275,7 @@ Size MatSize::operator()() const { int d = dims(); CV_DbgAssert(d <= 2); - return d == 2 ? Size(p[1], p[0]) : Size(p[0], 1); + return d == 2 ? Size(p[1], p[0]) : d == 1 ? Size(p[0], 1) : Size(0, 0); } inline ```
When empty matrix is created, dims()==0, but total() is 0 too. But there can be scalar matrix with dims() == 0, but total() == 1: ``` Mat m, mscalar(0, nullptr, CV_32F); mscalar.at<float>(0) = 1.f; printf("empty m dims: %d, total number of elements: %zu, empty m.empty(): %d, " "scalar m dims: %d, scalar m total: %zu, scalar m.empty(): %d\n", m.dims, m.total(), m.empty(), mscalar.dims, mscalar.total(), mscalar.empty()); ``` that is, in addition to dims member, empty() should also be used to distinguish between empty matrix and scalar matrix.
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/mat.inl.hpp **Change Type:** modified **Context:** PR #25945: Fix size() for 0d matrix **Review Line:** 1278 **Code Changes:** ```diff @@ -1275,7 +1275,7 @@ Size MatSize::operator()() const { int d = dims(); CV_DbgAssert(d <= 2); - return d == 2 ? Size(p[1], p[0]) : Size(p[0], 1); + return d == 2 ? Size(p[1], p[0]) : d == 1 ? Size(p[0], 1) : Size(0, 0); } inline ```
Oh, regarding the other question. `1x1x1x1x1x1` is not a scalar, because `dims()==6`.
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/include/opencv2/core/mat.inl.hpp **Change Type:** modified **Context:** PR #25945: Fix size() for 0d matrix **Review Line:** 1278 **Code Changes:** ```diff @@ -1275,7 +1275,7 @@ Size MatSize::operator()() const { int d = dims(); CV_DbgAssert(d <= 2); - return d == 2 ? Size(p[1], p[0]) : Size(p[0], 1); + return d == 2 ? Size(p[1], p[0]) : d == 1 ? Size(p[0], 1) : Size(0, 0); } inline ```
Thx for the explanation. It seems the scalar case is already handled properly because `d==1` in MatSize (though `dims()==0` in cv::Mat). I added a test for that.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
it could be simplified.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
Please remove VKCOM. It does not work for now.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
The same: cv.dnn.DNN_TARGET_VULKAN
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Review Line:** 158 **Code Changes:** ```diff + + def make_schedule(self, ddim_num_steps, ddim_discretize="uniform", ddim_eta=0., verbose=True): + self.ddim_timesteps = make_ddim_timesteps(ddim_discr_method=ddim_discretize, num_ddim_timesteps=ddim_num_steps, + num_ddpm_timesteps=self.ddpm_num_timesteps,verbose=verbose) + alphas_cumprod = self.model.alphas_cumprod + assert alphas_cumprod.shape[0] == self.ddpm_num_timesteps, 'alphas have to be defined for each timestep' + to_numpy = partial(np.array, copy=True, dtype=np.float32) + + self.register_buffer('betas', to_numpy(self.model.betas)) + self.register_buffer('alphas_cumprod', to_numpy(alphas_cumprod)) + self.register_buffer('alphas_cumprod_prev', to_numpy(self.model.alphas_cumprod_prev)) ```
I suppose there should be `num_ddpm_timesteps`. it's not used in the function. Also `self.ddpm_num_timesteps` is always 1000 as assigned in constructor and.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Review Line:** 475 **Code Changes:** ```diff + +def main(args): + + image = cv.imread(args.image) + mask = create_mask(deepcopy(image)) + image = cv.cvtColor(image, cv.COLOR_BGR2RGB) + + batch = make_batch(image, mask) + image, mask, masked_image = batch["image"], batch["mask"], batch["masked_image"] + + model = DDIMInpainter(args) ```
You can use new cv.IMREAD_COLOR_RGB for imread to get rid of cvtColor.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
Most of parameters are not used in sample. Do we really need them?
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Review Line:** 475 **Code Changes:** ```diff + +def main(args): + + image = cv.imread(args.image) + mask = create_mask(deepcopy(image)) + image = cv.cvtColor(image, cv.COLOR_BGR2RGB) + + batch = make_batch(image, mask) + image, mask, masked_image = batch["image"], batch["mask"], batch["masked_image"] + + model = DDIMInpainter(args) ```
I need to this seperately since `create_maks(...)` will show image with reversed channel if I use `cv.IMREAD_COLOR_RGB` in the `imread`
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
I would suggest put backends and targets that you tried. Others are not common.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
Prefer to use `np.newaxis` to make a new dimension, e.g. `image = image[np.newaxis, ...]`
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
Use `np.newaxis` as above
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
It seems `kwargs` is not needed.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
I believe it should be `np.newaxis`
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
There is no backend and target for openvino in line 46 to 47.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** samples/dnn/ldm_inpainting.py **Change Type:** added **Context:** PR #25950: Diffusion Inpainting Sample **Code Changes:** ```diff @@ -0,0 +1,489 @@ +import cv2 as cv +import numpy as np +import argparse +from tqdm import tqdm +from functools import partial +from copy import deepcopy + +## let use write description of the script and general information how to use it + ```
`np.newaxis`
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/js/generator/CMakeLists.txt **Change Type:** modified **Context:** PR #25986: Split Javascript white-list to support contrib modules **Code Changes:** ```diff @@ -38,8 +38,21 @@ set(scripts_hdr_parser "${JS_SOURCE_DIR}/../python/src2/hdr_parser.py") if(DEFINED ENV{OPENCV_JS_WHITELIST}) set(OPENCV_JS_WHITELIST_FILE "$ENV{OPENCV_JS_WHITELIST}") + message(STATUS "Use white list from environment ${OPENCV_JS_WHITELIST_FILE}") else() - set(OPENCV_JS_WHITELIST_FILE "${OpenCV_SOURCE_DIR}/platforms/js/opencv_js.config.py") + #generate white list from modules/<module_name>/misc/js/whitelist.json + set(OPENCV_JS_WHITELIST_FILE "${CMAKE_CURRENT_BINARY_DIR}/whitelist.json") + foreach(m in ${OPENCV_JS_MODULES}) ```
No need to write file each time. Generate string and update on changes only. See `ocv_update_file` Also generated file should be added as a dependency to the build graph of bindings generator target.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/js/generator/embindgen.py **Change Type:** modified **Context:** PR #25986: Split Javascript white-list to support contrib modules **Review Line:** 956 **Code Changes:** ```diff @@ -76,6 +76,7 @@ else: from cStringIO import StringIO +import json func_table = {} @@ -103,11 +104,32 @@ def makeWhiteList(module_list): wl[k] = m[k] return wl +def makeWhiteListJson(module_list): + wl = {} + for n, gen_dict in module_list.items(): ```
Old behavior seems broken. Besides of json there are namespaces configuration were possible (see `namespace_prefix_override`), so python configuration code is preferable and provides full customization without unnecessary limitation of json.
You are an expert OpenCV code reviewer specializing in code quality and best practices. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/js/generator/CMakeLists.txt **Change Type:** modified **Context:** PR #25986: Split Javascript white-list to support contrib modules **Code Changes:** ```diff @@ -38,8 +38,21 @@ set(scripts_hdr_parser "${JS_SOURCE_DIR}/../python/src2/hdr_parser.py") if(DEFINED ENV{OPENCV_JS_WHITELIST}) set(OPENCV_JS_WHITELIST_FILE "$ENV{OPENCV_JS_WHITELIST}") + message(STATUS "Use white list from environment ${OPENCV_JS_WHITELIST_FILE}") else() - set(OPENCV_JS_WHITELIST_FILE "${OpenCV_SOURCE_DIR}/platforms/js/opencv_js.config.py") + #generate white list from modules/<module_name>/misc/js/whitelist.json + set(OPENCV_JS_WHITELIST_FILE "${CMAKE_CURRENT_BINARY_DIR}/whitelist.json") + foreach(m in ${OPENCV_JS_MODULES}) ```
ocv_update_file - done. dependency is already there.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/perf/perf_layer.cpp **Change Type:** modified **Context:** PR #23279: dnn: add ONNX TopK **Review Line:** 1097 **Code Changes:** ```diff +} +PERF_TEST_P_(Layer_TopK, TopK_3D_Axis1) { + test_layer(input_shape_3d, input_shape_3d[1] / 2, 1); +} +PERF_TEST_P_(Layer_TopK, TopK_3D_Axis2) { + test_layer(input_shape_3d, input_shape_3d[2] / 2, 2); +} +INSTANTIATE_TEST_CASE_P(/**/, Layer_TopK, + dnnBackendsAndTargets(/* withInferenceEngine= */ false, + /* withHalide= */ false, + /* withCpuOCV= */ true, ```
I propose to add cases for top-5 at least. It's popular case for classification scenarios.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/imgproc/test/test_color.cpp **Change Type:** modified **Context:** PR #25968: Correct Bayer2Gray u8 SIMD **Code Changes:** ```diff @@ -1863,6 +1863,26 @@ TEST(Imgproc_ColorBayer, regression) EXPECT_EQ(0, countNonZero(diff.reshape(1) > 1)); } +TEST(Imgproc_ColorBayer2Gray, regression_25823) +{ + const int n = 100; + Mat src(n, n, CV_8UC1); + Mat dst; + ```
``` cv::Mat gt(rows, cols, cv::Scalar(1)); EXPECT_EQ(0, cv::norm(dst, gt, NORM_INF)); ```
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/cuda4dnn/csl/tensor.hpp **Change Type:** modified **Context:** PR #26043: Support Matrices with axes > 6 CUDA **Review Line:** 31 **Code Changes:** ```diff @@ -28,7 +28,7 @@ #include <utility> #ifndef CSL_MAX_TENSOR_RANK - #define CSL_MAX_TENSOR_RANK 6 + #define CSL_MAX_TENSOR_RANK 7 #endif namespace cv { namespace dnn { namespace cuda4dnn { namespace csl { ```
Not sure why there is a rank limit. Could it be removed?
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/cuda4dnn/csl/tensor.hpp **Change Type:** modified **Context:** PR #26043: Support Matrices with axes > 6 CUDA **Review Line:** 31 **Code Changes:** ```diff @@ -28,7 +28,7 @@ #include <utility> #ifndef CSL_MAX_TENSOR_RANK - #define CSL_MAX_TENSOR_RANK 6 + #define CSL_MAX_TENSOR_RANK 7 #endif namespace cv { namespace dnn { namespace cuda4dnn { namespace csl { ```
I took a look at it. It seems that currently this variables is deeply rooted in CUDA backend. We can remove it from assertions and nothing happens. But to remove it completed quite a number of classes and function need to be refactored.
You are an expert OpenCV code reviewer specializing in header file design and interface. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/test/test_onnx_conformance_layer_filter__cuda_denylist.inl.hpp **Change Type:** modified **Context:** PR #26043: Support Matrices with axes > 6 CUDA **Review Line:** 93 **Code Changes:** ```diff @@ -90,7 +90,6 @@ "test_scatternd_max", "test_scatternd_min", "test_scatternd_multiply", -"test_nllloss_NCd1d2d3d4d5_none_no_weight_expanded", // crash: https://github.com/opencv/opencv/issues/25471 "test_dequantizelinear_blocked", // Issue https://github.com/opencv/opencv/issues/25999 "test_quantizelinear", // Issue https://github.com/opencv/opencv/issues/25999 "test_quantizelinear_axis", // Issue https://github.com/opencv/opencv/issues/25999 ```
just remove the line
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/test/test_onnx_importer.cpp **Change Type:** modified **Context:** PR #26073: (5.x) Merge 4.x - Contrib: https://github.com/opencv/opencv_contrib/pull/3783 Extra: https://github.com/opencv/opencv_extra/pull/1205 #2327... **Review Line:** 3302 **Code Changes:** ```diff + net.forward(outputs, std::vector<std::string>{"values", "indices"}); + + Mat output_res_val = outputs.front(), + output_res_ind = outputs.back(); + + output_ref_ind.convertTo(output_ref_ind, CV_32F); // TODO: revise this conversion in 5.x + + normAssert(output_ref_val, output_res_val, (basename + " values").c_str(), l1 ? l1 : default_l1, lInf ? lInf : default_lInf); + normAssert(output_ref_ind, output_res_ind, (basename + " indices").c_str(), l1 ? l1 : default_l1, lInf ? lInf : default_lInf); + + expectNoFallbacksFromIE(net); ```
Well, it needs to be handled in `getType` in the layer impementation but 4.x does not have this. I can prepare a dedicated one for the 5.x merge.
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/test/test_onnx_importer.cpp **Change Type:** modified **Context:** PR #26073: (5.x) Merge 4.x - Contrib: https://github.com/opencv/opencv_contrib/pull/3783 Extra: https://github.com/opencv/opencv_extra/pull/1205 #2327... **Review Line:** 3302 **Code Changes:** ```diff + net.forward(outputs, std::vector<std::string>{"values", "indices"}); + + Mat output_res_val = outputs.front(), + output_res_ind = outputs.back(); + + output_ref_ind.convertTo(output_ref_ind, CV_32F); // TODO: revise this conversion in 5.x + + normAssert(output_ref_val, output_res_val, (basename + " values").c_str(), l1 ? l1 : default_l1, lInf ? lInf : default_lInf); + normAssert(output_ref_ind, output_res_ind, (basename + " indices").c_str(), l1 ? l1 : default_l1, lInf ? lInf : default_lInf); + + expectNoFallbacksFromIE(net); ```
Filed ticket about it: https://github.com/opencv/opencv/issues/26076
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/core/test/test_operations.cpp **Change Type:** modified **Context:** PR #26073: (5.x) Merge 4.x - Contrib: https://github.com/opencv/opencv_contrib/pull/3783 Extra: https://github.com/opencv/opencv_extra/pull/1205 #2327... **Code Changes:** ```diff @@ -42,7 +42,7 @@ #include "test_precomp.hpp" #include "opencv2/ts/ocl_test.hpp" // T-API like tests -#include "opencv2/core/core_c.h" +#include <fenv.h> namespace opencv_test { namespace { @@ -1087,7 +1087,6 @@ bool CV_OperationsTest::operations1() ```
double check for what?
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/einsum_layer.cpp **Change Type:** modified **Context:** PR #26073: (5.x) Merge 4.x - Contrib: https://github.com/opencv/opencv_contrib/pull/3783 Extra: https://github.com/opencv/opencv_extra/pull/1205 #2327... **Review Line:** 545 **Code Changes:** ```diff @@ -541,7 +542,7 @@ class LayerEinsumImpl CV_FINAL : public EinsumLayer // Use either the preprocessed inputs (if it is available) or the corresponding raw inputs result = pairwiseOperandProcess(!result.empty() ? result : rawInputs[0], !result.empty() ? tmpResult : homogenizedInputDims[0], - (!preProcessedInputs.empty() && !preProcessedInputs[input].empty()) ? preProcessedInputs[input] : rawInputs[input], + (!preProcessedInputs[input].empty()) ? preProcessedInputs[input] : rawInputs[input], homogenizedInputDims[input], reducedDims, isFinalPair); @@ -605,8 +606,8 @@ void LayerEinsumImpl::preProcessInputs(InputArrayOfArrays& inputs_arr) std::vector<cv::Mat> inputs; ```
Why is pre-check removed? (anyway it is wrong as it doesn't check vector size vs index)
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/einsum_layer.cpp **Change Type:** modified **Context:** PR #26073: (5.x) Merge 4.x - Contrib: https://github.com/opencv/opencv_contrib/pull/3783 Extra: https://github.com/opencv/opencv_extra/pull/1205 #2327... **Review Line:** 542 **Code Changes:** ```diff + CV_CheckEQ((size_t)inputs_arr.total(), (size_t)numInputs, "Number of inputs in forward and inputs during graph constructions do not match"); if (inputs_arr.depth() == CV_16F) { @@ -541,7 +542,7 @@ class LayerEinsumImpl CV_FINAL : public EinsumLayer // Use either the preprocessed inputs (if it is available) or the corresponding raw inputs result = pairwiseOperandProcess(!result.empty() ? result : rawInputs[0], !result.empty() ? tmpResult : homogenizedInputDims[0], - (!preProcessedInputs.empty() && !preProcessedInputs[input].empty()) ? preProcessedInputs[input] : rawInputs[input], + (!preProcessedInputs[input].empty()) ? preProcessedInputs[input] : rawInputs[input], homogenizedInputDims[input], ```
Lost assertion: https://github.com/opencv/opencv/blame/4.x/modules/dnn/src/layers/einsum_layer.cpp#L462
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/src/layers/einsum_layer.cpp **Change Type:** modified **Context:** PR #26073: (5.x) Merge 4.x - Contrib: https://github.com/opencv/opencv_contrib/pull/3783 Extra: https://github.com/opencv/opencv_extra/pull/1205 #2327... **Review Line:** 545 **Code Changes:** ```diff @@ -541,7 +542,7 @@ class LayerEinsumImpl CV_FINAL : public EinsumLayer // Use either the preprocessed inputs (if it is available) or the corresponding raw inputs result = pairwiseOperandProcess(!result.empty() ? result : rawInputs[0], !result.empty() ? tmpResult : homogenizedInputDims[0], - (!preProcessedInputs.empty() && !preProcessedInputs[input].empty()) ? preProcessedInputs[input] : rawInputs[input], + (!preProcessedInputs[input].empty()) ? preProcessedInputs[input] : rawInputs[input], homogenizedInputDims[input], reducedDims, isFinalPair); @@ -605,8 +606,8 @@ void LayerEinsumImpl::preProcessInputs(InputArrayOfArrays& inputs_arr) std::vector<cv::Mat> inputs; ```
After https://github.com/opencv/opencv/pull/26059 the `preProcessedInputs` is resized, but not reserved. As soon as the layer has inputs, the vector is not empty. I just selected 4.x version.
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/imgproc/perf/perf_warp.cpp **Change Type:** modified **Context:** PR #26054: Extended perf tests for warpPerspective to cover channels too **Review Line:** 45 **Code Changes:** ```diff TEST_CYCLE() warpAffine( src, dst, warpMat, sz, interType, borderMode, borderColor ); -#ifdef __ANDROID__ - SANITY_CHECK(dst, interType==INTER_LINEAR? 5 : 10); -#else SANITY_CHECK(dst, 1); -#endif } PERF_TEST_P(TestWarpAffine, DISABLED_WarpAffine_ovx, @@ -72,29 +68,28 @@ PERF_TEST_P(TestWarpAffine, DISABLED_WarpAffine_ovx, ```
Maybe `SANITY_CHECK_NOTHING()`? We're trying to deprecate accuracy checks in perf tests for some time (https://github.com/opencv/opencv/wiki/HowToWritePerfTests#simple-regression-check).
You are an expert OpenCV code reviewer specializing in implementation quality. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/imgproc/perf/perf_warp.cpp **Change Type:** modified **Context:** PR #26054: Extended perf tests for warpPerspective to cover channels too **Review Line:** 45 **Code Changes:** ```diff TEST_CYCLE() warpAffine( src, dst, warpMat, sz, interType, borderMode, borderColor ); -#ifdef __ANDROID__ - SANITY_CHECK(dst, interType==INTER_LINEAR? 5 : 10); -#else SANITY_CHECK(dst, 1); -#endif } PERF_TEST_P(TestWarpAffine, DISABLED_WarpAffine_ovx, @@ -72,29 +68,28 @@ PERF_TEST_P(TestWarpAffine, DISABLED_WarpAffine_ovx, ```
Let's presume the checks, where we have it. No check for new code.
You are an expert OpenCV code reviewer specializing in testing and validation. Review the provided code changes and provide specific, actionable feedback.
**File:** modules/dnn/misc/python/test/test_dnn.py **Change Type:** modified **Context:** PR #26026: Add support for boolan input/outputs in python bindings **Code Changes:** ```diff @@ -546,5 +546,24 @@ def test_net_builder(self): out = net.forward() self.assertEqual(out.shape, (1, 2, 3, 4)) + def test_bool_operator(self): + n = self.find_dnn_file('dnn/onnx/models/and_op.onnx') + + x = np.random.randint(0, 2, [5], dtype=np.bool_) + y = np.random.randint(0, 2, [5], dtype=np.bool_) + o = x & y ```
Can we generate test data in runtime instead?