GSoC Add ONNX Support for GatherElements by Aser-Abdelfatah · Pull Request #24092 · opencv/opencv

Aser-Abdelfatah · 2023-08-02T00:12:38Z

Pull Request Readiness Checklist

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

I agree to contribute to the project under Apache 2 License.
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
The PR is proposed to the proper branch
There is a reference to the original bug report and related work
There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
The feature is well documented and sample code can be built with the project CMake

Adds support to the ONNX operator GatherElements operator docs
Added tests to opencv_extra at pull request opencv/opencv_extra#1082

modules/dnn/include/opencv2/dnn/all_layers.hpp

modules/dnn/test/test_onnx_importer.cpp

modules/dnn/src/onnx/onnx_importer.cpp

fengyuentau

Good work! Please read reviewers' comments and give responses.

modules/dnn/src/layers/gather_elements_layer.cpp

dkurt · 2023-08-03T19:30:44Z

@Aser-Abdelfatah , what about PR #23975? Can be closed?

fengyuentau

Other looks good to me! 👍 Good job!

modules/dnn/src/onnx/onnx_importer.cpp

modules/dnn/src/layers/gather_elements_layer.cpp

modules/dnn/test/test_onnx_importer.cpp

fengyuentau · 2023-09-14T01:37:52Z

Hello @Aser-Abdelfatah , could you solve the patch size limitation problem with using existing images from https://github.com/opencv/opencv_extra/tree/4.x/testdata/gpu/opticalflow?

Aser-Abdelfatah · 2023-09-19T01:46:45Z

Hi,
Sorry for getting back late to this. I was looking for a way to convert the pre-processed image data to npy file using C++. In this OpenCV forum thread that I opened https://forum.opencv.org/t/how-to-export-an-opencv-mat-to-npy-pb-file-in-c/14464/4, it's mentioned that the size of the data will be the same, which makes perfect sense, since we are storing the same data in the same format.

dkurt · 2023-09-19T03:28:06Z

@Aser-Abdelfatah , there is no need to create npy in C++. Inputs are images so use imread and blobFromImage as mentioned here: opencv/opencv_extra#1082 (comment)

fengyuentau · 2023-09-19T03:59:33Z

I think @Aser-Abdelfatah is going to reuse testONNXModels right? In case of using real images as inputs, I think you should not use testONNXModels, but can refer to (skip the part of creating reference data)

opencv/modules/dnn/test/test_onnx_importer.cpp

Lines 2470 to 2494 in 157b0e7

    
           TEST_P(Test_ONNX_layers, YOLOv7) 
        
           { 
        
               std::string weightPath = _tf("models/yolov7_not_simplified.onnx", false); 
        
               std::string imgPath = _tf("../dog_orig_size.png"); 
        
               Size targetSize{640, 640}; 
        
               float conf_threshold = 0.3; 
        
               float iou_threshold = 0.5; 
        
               // Reference, which is collected with input size of 640x640 
        
               std::vector<int> refClassIds{1, 16, 7}; 
        
               std::vector<float> refScores{0.9614331f, 0.9589417f, 0.8679074f}; 
        
               // [x1, y1, x2, y2] x 3 
        
               std::vector<Rect2d> refBoxes{Rect2d(105.973236f, 150.16716f,  472.59012f, 466.48834f), 
        
                                             Rect2d(109.97953f,  246.17862f, 259.83676f, 600.76624f), 
        
                                             Rect2d(385.96185f, 83.02809f,  576.07355f,  189.82793f)}; 
        
               Mat img = imread(imgPath); 
        
               Mat inp = blobFromImage(img, 1/255.0, targetSize, Scalar(0, 0, 0), true, false); 
        
               Net net = readNet(weightPath); 
        
               net.setInput(inp); 
        
               std::vector<Mat> outs; 
        
               net.forward(outs, net.getUnconnectedOutLayersNames());

. You will need to run with ONNX Runtime to save reference output first, then load to ref mat with blobFromNPY.

vpisarev · 2023-09-27T06:08:18Z

modules/dnn/src/layers/gather_elements_layer.cpp

+            const T* tmp_p_data = p_data + inp_offset;
+            T* tmp_p_out = p_out + ind_offset;
+            for (size_t i = 0; i < total; i++) {
+                *tmp_p_out = *tmp_p_data;


sounds like a bug here

fengyuentau · 2023-09-29T08:43:58Z

The raft model has a GatherElements operator which has a non-constant indices input. ~~We need it constant so as to allocate memory in pre-run stage. This operator is in between conv layers, most probably not pre- or post-processing and cannot be removed.~~ It may work even with non-constant indices if its shape is known.

dkurt · 2023-09-29T11:37:50Z

modules/dnn/test/test_onnx_importer.cpp

+    net.setInput(blob0, "0");
+    net.setInput(blob1, "1");
+    auto out0 = net.forward("12007");
+    auto out1 = net.forward("12006");


Forward model only once and request both outputs

dkurt · 2023-09-29T11:40:36Z

modules/dnn/test/test_onnx_importer.cpp

+    std::string img0_path = findDataFile(std::string("gpu/opticalflow/frame0.png"));
+    std::string img1_path = findDataFile(std::string("gpu/opticalflow/frame1.png"));
+
+    Size target_size{480, 360};


@fengyuentau, can we reduce input resolution as minimul as possible to cover required checks for a new layer? 240x180 or lower.

It may be feasible, but the model may be sensitive to the input shape. The intention was we use the model in opencv_zoo for accuracy or performance testing. If the model is shape-sensitive in the end, we have to replace the model in the zoo with one with a smaller input shape. That could be some trouble.

Can you please reduce precision to FP16? Some changes in blobFromNPY will be required.

I will handle this problem after we can run this model. Currently we are facing some other issues running this model.

How about we just check the smaller output? This model has two outputs, one of which has shape [1, 2, 45, 60] and file size 21KB, the other of which has shape [1, 2, 360, 480] and file size 1.32MB. The bigger output is calculated from the smaller output. I think it also makes sense if we just check the smaller one. In this case, we can avoid replacing models.

Finally dnn can achieve the same accuracy with #24386 (greater fix), with or without #24378 (instancenorm).

This RAFT model takes nearly 2s for a single inference on my mac m1. I think we still need to replace with a smaller input scale version.

Tested with a smaller input one but the time is almost the same. Lets keep what we have. Also I found that we have almost the same inference time compared to ONNX Runtime.

I suggest to add a tag to this test applyTestTag(CV_TEST_TAG_LONG, CV_TEST_TAG_DEBUG_VERYLONG, CV_TEST_TAG_MEMORY_2GB); to skip as long. PR is ready and might be merged.

I suggest to add a tag to this test applyTestTag(CV_TEST_TAG_LONG, CV_TEST_TAG_DEBUG_VERYLONG, CV_TEST_TAG_MEMORY_2GB); to skip as long. PR is ready and might be merged.

Thank you for the tips. I almost forgot these flags!

fengyuentau · 2023-10-07T08:18:03Z

modules/dnn/src/layers/gather_elements_layer.cpp

+    }
+
+    template<typename T>
+    void forward_impl(const Mat& data, const Mat& indices,  Mat& out)


We need to refactor this implementation because it is just tooooooo slow, leading to a very long inference with the RAFT model (it has 160!!! GatherElements operators). Perf with GatherElements(data:FLOAT[2700, 1, 2914], indices:INT64[2700, 1, 81]):
ort_benchmark.zip

this impl ort

15415.07 ms 0.15ms

Improvement:

Previous impl by Aser: 15415.07 ms
After removing meaningless for loop: 1ms
Speedup: 0.49ms
ort: 0.15 ms

dnn: fix inconsistent input dtype for nary eltwise layers #24386 Resolves #24385 Merge with opencv/opencv_extra#1107 Relates #24092 (comment) ### 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

… to Test_ONNX_nets

…e tests

Signed-off-by: Yuantao Feng <yuantao.feng@opencv.org.cn>

asmorkalov · 2023-10-13T09:42:05Z

@vpisarev @dkurt Could you take a look on the last changes?

modules/dnn/test/test_onnx_importer.cpp

Signed-off-by: Yuantao Feng <yuantao.feng@opencv.org.cn>

…ents_ONNX Merge with opencv/opencv#24092 This pull request serves as an addition to the pull request "GSoC Add ONNX Support for GatherElements #24092", which is submitted to the original OpenCV repository. It contains three tests to the GatherElements operator. The three tests are sourced from: * https://github.com/onnx/onnx/tree/main/onnx/backend/test/data/node/test_gather_elements_0 * https://github.com/onnx/onnx/tree/main/onnx/backend/test/data/node/test_gather_elements_1 * https://github.com/onnx/onnx/tree/main/onnx/backend/test/data/node/test_gather_elements_negative_indices And the models were configured using the code provided by Yuantao Feng at https://github.com/fengyuentau/onnx_utils to confirm output shape in pre-run. --- Reference outputs for `optical_flow_estimation_raft_2023aug.onnx` is generated with the following script: ```python import cv2 as cv import numpy as np img1 = cv.imread("./opencv_extra/testdata/gpu/opticalflow/frame0.png") img2 = cv.imread("./opencv_extra/testdata/gpu/opticalflow/frame1.png") img1_blob = cv.cvtColor(img1, cv.COLOR_BGR2RGB) img1_blob = cv.resize(img1_blob, (480, 360)) # source: (1, 3, 388, 584) # target: 1, 3, 360, 480 img1_blob = cv.dnn.blobFromImage(img1_blob) # source: (1, 3, 388, 584) # target: 1, 3, 360, 480 img2_blob = cv.cvtColor(img2, cv.COLOR_BGR2RGB) img2_blob = cv.resize(img2_blob, (480, 360)) img2_blob = cv.dnn.blobFromImage(img2_blob) print(img1_blob.shape, img2_blob.shape) import onnxruntime as ort net = ort.InferenceSession("optical_flow_estimation_raft_2023aug.onnx", providers=["CPUExecutionProvider"]) output = net.run(["12007", "12006"], {"0": img1_blob, "1": img2_blob}) np.save("output_optical_flow_estimation_raft_2023aug_0.npy", output[0]) np.save("output_optical_flow_estimation_raft_2023aug_1.npy", output[1]) print(output[0].shape) print(output[1].shape) ```

asmorkalov · 2023-10-17T15:22:48Z

@opencv-alalek please update models on Buildbot.

opencv-alalek · 2023-10-19T04:54:28Z

modules/dnn/src/layers/gather_elements_layer.cpp

+        CV_TRACE_FUNCTION();
+        CV_TRACE_ARG_VALUE(name, "name", name.c_str());
+
+        std::vector<Mat> inputs, outputs;


Implementation should handle FP16 datatype too

[ FAILED ] 3 tests, listed below: [ FAILED ] Test_ONNX_conformance.Layer_Test/test_gather_elements_0_OCV_OCL_FP16, where GetParam() = (test_gather_elements_0, OCV/OCL_FP16) [ FAILED ] Test_ONNX_conformance.Layer_Test/test_gather_elements_1_OCV_OCL_FP16, where GetParam() = (test_gather_elements_1, OCV/OCL_FP16) [ FAILED ] Test_ONNX_conformance.Layer_Test/test_gather_elements_negative_indices_OCV_OCL_FP16, where GetParam() = (test_gather_elements_negative_indices, OCV/OCL_FP16)

https://pullrequest.opencv.org/buildbot/builders/4_x-lin64/builds/100283

Ok, I will do this.

opencv-alalek · 2023-10-19T04:56:49Z

@asmorkalov Why is it merged with failed pre-commit?

https://pullrequest.opencv.org/buildbot/builders/precommit_opencl_linux/builds/100255

dnn onnx: add instance norm layer #24378 Resolves #24377 Relates #24092 (comment) | Perf | multi-thread | single-thread | | - | - | - | | x: [2, 64, 180, 240] | 3.95ms | 11.12ms | Todo: - [x] speed up by multi-threading - [x] add perf - [x] add backend: OpenVINO - [x] add backend: CUDA - [x] add backend: OpenCL (no fp16) - [ ] add backend: CANN (will be done via #24462) ### 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 ``` force_builders=Linux OpenCL,Win64 OpenCL,Custom buildworker:Custom=linux-4 build_image:Custom=ubuntu:18.04 modules_filter:Custom=none disable_ipp:Custom=ON ```

dnn: fix inconsistent input dtype for nary eltwise layers opencv#24386 Resolves opencv#24385 Merge with opencv/opencv_extra#1107 Relates opencv#24092 (comment) ### 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

…herElements_ONNX GSoC Add ONNX Support for GatherElements opencv#24092 Merge with: opencv/opencv_extra#1082 Adds support to the ONNX operator GatherElements [operator docs](https://github.com/onnx/onnx/blob/main/docs/Operators.md#GatherElements) Added tests to opencv_extra at pull request opencv/opencv_extra#1082 ### 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

dnn onnx: add instance norm layer opencv#24378 Resolves opencv#24377 Relates opencv#24092 (comment) | Perf | multi-thread | single-thread | | - | - | - | | x: [2, 64, 180, 240] | 3.95ms | 11.12ms | Todo: - [x] speed up by multi-threading - [x] add perf - [x] add backend: OpenVINO - [x] add backend: CUDA - [x] add backend: OpenCL (no fp16) - [ ] add backend: CANN (will be done via opencv#24462) ### 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 ``` force_builders=Linux OpenCL,Win64 OpenCL,Custom buildworker:Custom=linux-4 build_image:Custom=ubuntu:18.04 modules_filter:Custom=none disable_ipp:Custom=ON ```

dnn: fix inconsistent input dtype for nary eltwise layers opencv#24386 Resolves opencv#24385 Merge with opencv/opencv_extra#1107 Relates opencv#24092 (comment) ### 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

…herElements_ONNX GSoC Add ONNX Support for GatherElements opencv#24092 Merge with: opencv/opencv_extra#1082 Adds support to the ONNX operator GatherElements [operator docs](https://github.com/onnx/onnx/blob/main/docs/Operators.md#GatherElements) Added tests to opencv_extra at pull request opencv/opencv_extra#1082 ### 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

dnn onnx: add instance norm layer opencv#24378 Resolves opencv#24377 Relates opencv#24092 (comment) | Perf | multi-thread | single-thread | | - | - | - | | x: [2, 64, 180, 240] | 3.95ms | 11.12ms | Todo: - [x] speed up by multi-threading - [x] add perf - [x] add backend: OpenVINO - [x] add backend: CUDA - [x] add backend: OpenCL (no fp16) - [ ] add backend: CANN (will be done via opencv#24462) ### 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 ``` force_builders=Linux OpenCL,Win64 OpenCL,Custom buildworker:Custom=linux-4 build_image:Custom=ubuntu:18.04 modules_filter:Custom=none disable_ipp:Custom=ON ```

dnn: fix inconsistent input dtype for nary eltwise layers opencv#24386 Resolves opencv#24385 Merge with opencv/opencv_extra#1107 Relates opencv#24092 (comment) ### 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

…herElements_ONNX GSoC Add ONNX Support for GatherElements opencv#24092 Merge with: opencv/opencv_extra#1082 Adds support to the ONNX operator GatherElements [operator docs](https://github.com/onnx/onnx/blob/main/docs/Operators.md#GatherElements) Added tests to opencv_extra at pull request opencv/opencv_extra#1082 ### 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

dnn onnx: add instance norm layer opencv#24378 Resolves opencv#24377 Relates opencv#24092 (comment) | Perf | multi-thread | single-thread | | - | - | - | | x: [2, 64, 180, 240] | 3.95ms | 11.12ms | Todo: - [x] speed up by multi-threading - [x] add perf - [x] add backend: OpenVINO - [x] add backend: CUDA - [x] add backend: OpenCL (no fp16) - [ ] add backend: CANN (will be done via opencv#24462) ### 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 ``` force_builders=Linux OpenCL,Win64 OpenCL,Custom buildworker:Custom=linux-4 build_image:Custom=ubuntu:18.04 modules_filter:Custom=none disable_ipp:Custom=ON ```

dkurt reviewed Aug 2, 2023

View reviewed changes

modules/dnn/include/opencv2/dnn/all_layers.hpp Show resolved Hide resolved

dkurt reviewed Aug 2, 2023

View reviewed changes

modules/dnn/test/test_onnx_importer.cpp Outdated Show resolved Hide resolved

dkurt reviewed Aug 2, 2023

View reviewed changes

modules/dnn/src/onnx/onnx_importer.cpp Outdated Show resolved Hide resolved

dkurt reviewed Aug 2, 2023

View reviewed changes

modules/dnn/src/onnx/onnx_importer.cpp Show resolved Hide resolved

fengyuentau reviewed Aug 2, 2023

View reviewed changes

modules/dnn/src/layers/gather_elements_layer.cpp Outdated Show resolved Hide resolved

modules/dnn/src/layers/gather_elements_layer.cpp Outdated Show resolved Hide resolved

fengyuentau added the GSoC label Aug 2, 2023

fengyuentau added category: dnn category: dnn (onnx) ONNX suport issues in DNN module labels Aug 4, 2023

fengyuentau reviewed Aug 25, 2023

View reviewed changes

modules/dnn/src/onnx/onnx_importer.cpp Show resolved Hide resolved

modules/dnn/src/layers/gather_elements_layer.cpp Outdated Show resolved Hide resolved

fengyuentau self-assigned this Aug 25, 2023

fengyuentau added this to the 4.9.0 milestone Aug 25, 2023

fengyuentau reviewed Aug 25, 2023

View reviewed changes

modules/dnn/test/test_onnx_importer.cpp Outdated Show resolved Hide resolved

dkurt reviewed Aug 29, 2023

View reviewed changes

modules/dnn/test/test_onnx_importer.cpp Outdated Show resolved Hide resolved

fengyuentau mentioned this pull request Sep 5, 2023

GSoC Add ONNX Support for GatherElements opencv/opencv_extra#1082

Merged

vpisarev reviewed Sep 27, 2023

View reviewed changes

dkurt reviewed Sep 29, 2023

View reviewed changes

fengyuentau reviewed Oct 7, 2023

View reviewed changes

This was referenced Oct 9, 2023

dnn: inference results are different from ONNX Runtime on the InstanceNorm layer #24377

Closed

dnn onnx: add instance norm layer #24378

Merged

dnn: fix inconsistent input dtype for nary eltwise layers #24386

Merged

Aser-Abdelfatah added 2 commits October 13, 2023 17:30

Add Support for GatherElements

86bd690

response to comments

4fee29e

fengyuentau added 5 commits October 13, 2023 17:30

update to use multi threading

d949dbd

forward once; only check the smaller output; replace RAFT test domain…

e2fec12

… to Test_ONNX_nets

remove gather elements tests in test_onnx_importer; enable conformanc…

8d3be8e

…e tests

sync ref file name; adjust threshold

2dd7b21

Signed-off-by: Yuantao Feng <yuantao.feng@opencv.org.cn>

add gather elements to vulkan filter list

6c3a8a1

enable raft test

ac3ea42

fengyuentau force-pushed the GSoC_Support_GatherElements_ONNX branch from 47901d3 to ac3ea42 Compare October 13, 2023 09:46

dkurt reviewed Oct 13, 2023

View reviewed changes

modules/dnn/test/test_onnx_importer.cpp Outdated Show resolved Hide resolved

fengyuentau added 2 commits October 16, 2023 10:15

tune threshold

e8e1637

Signed-off-by: Yuantao Feng <yuantao.feng@opencv.org.cn>

set skip flags

83031aa

dkurt approved these changes Oct 16, 2023

View reviewed changes

asmorkalov merged commit 240b245 into opencv:4.x Oct 18, 2023

dkurt mentioned this pull request Oct 18, 2023

Update DNN Optical Flow sample with RAFT model #24424

Closed

opencv-alalek reviewed Oct 19, 2023

View reviewed changes

fengyuentau mentioned this pull request Oct 19, 2023

dnn: fp16 support for gather elements #24427

Merged

6 tasks

asmorkalov mentioned this pull request Nov 3, 2023

(5.x) Merge 4.x #24486

Merged

Uh oh!

Conversation

Aser-Abdelfatah commented Aug 2, 2023 • edited by fengyuentau Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request Readiness Checklist

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

fengyuentau left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

dkurt commented Aug 3, 2023

Uh oh!

fengyuentau left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

fengyuentau commented Sep 14, 2023

Uh oh!

Aser-Abdelfatah commented Sep 19, 2023

Uh oh!

dkurt commented Sep 19, 2023

Uh oh!

fengyuentau commented Sep 19, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fengyuentau commented Sep 29, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fengyuentau Oct 7, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fengyuentau Oct 7, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

asmorkalov commented Oct 13, 2023

Uh oh!

Uh oh!

asmorkalov commented Oct 17, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Aser-Abdelfatah commented Aug 2, 2023 •

edited by fengyuentau

Loading

fengyuentau commented Sep 29, 2023 •

edited

Loading

fengyuentau Oct 7, 2023 •

edited

Loading

fengyuentau Oct 7, 2023 •

edited

Loading