Int8 block layout support #28741
After this patch we got the following speed ups on **resnet50-qdq.onnx** model.
- Inference time now: **_~6.6ms_** (inference time using onnxruntime is ~5.7ms).
- Inference time before: _**~11.5ms**_ [after QDQ PR #28595]
- Speed up: _**~42.6% or 1.74x**_
- Device details:
- Model name: Intel(R) Core(TM) i9-14900KS, x86, 32 Cores, ubuntu 24.04,
### 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
Added Output Tensor Names support in new DNN engine #28637
closes: https://github.com/opencv/opencv/issues/26201
### 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
Block layout-based convolution in DNN #28585
merge together with https://github.com/opencv/opencv_extra/pull/1321
Some core parts of the new engine in DNN module have been revised substantially:
1. all tests seem to pass, except for `Test_Graph_Simplifier.ResizeSubgraph`, which has been disabled because it does not take the newly added `TransformLayoutLayer` into account. The test should be reworked perhaps.
1. convolution and related operations (maxpool/avgpool) now use so-called block layout (`DATA_LAYOUT_BLOCK`), where `NxCxHxW` tensors are represented as `NxC1xHxWxC0`, where `C1=(C + C0-1)/C0` and `C0` is a power-of-two (usually 4, 8, 16 or 32).
1. graph is now pre-processed and `TransformLayoutLayer` is inserted to convert data from NCHW or NHWC layout to the block layout or vice versa. The transformations are done in a lazy way only when they are really needed. For example, in the whole Resnet only 2 transformations are performed.
1. transformer-based models and other models that do not use convolutions will run as usual, without going to block layout.
1. there is yet another graph preprocessing stage added that embeds constant weights/scale and bias into convolution and batch norm layers.
1. 'batchnorm', 'activation' and 'adding a residual' are now fused with convolution, just like in the old engine. That brings some noticeable acceleration.
1. optimized convolution kernels have been added.
* depthwise convolution, as well as maxpool and avgpool support C0=4, 8, 16 etc. _as long as_ C0 is divisible by the number of fp32 lanes in a SIMD register of the target platform (e.g. on ARM with NEON there must be `C0 % 4 == 0`, on x64 with AVX2 `C0 % 8 == 0`).
* non-depthwise convolution only supports C0=8 for now. C0=8 seems to be a sweetspot for ARM with NEON, x64 with AVX2 or RISC-V with RVV (with 128- or 256-bit registers). For some platforms with dedicated matrix accelerators C0=16 or even C0=32 might be more efficient, but we could add the respective kernels later.
* only fp32 kernels have been added. fp16/bf16 kernels might be added a little later.
### Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
- [x] Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake
New dnn engine #26056
This is the 1st PR with the new engine; CI is green and PR is ready to be merged, I think.
Merge together with https://github.com/opencv/opencv_contrib/pull/3794
---
**Known limitations:**
* [solved] OpenVINO is temporarily disabled, but is probably easy to restore (it's not a deal breaker to merge this PR, I guess)
* The new engine does not support any backends nor any targets except for the default CPU implementation. But it's possible to choose the old engine when loading a model, then all the functionality is available.
* [Caffe patch is here: #26208] The new engine only supports ONNX. When a model is constructed manually or is loaded from a file of different format (.tf, .tflite, .caffe, .darknet), the old engine is used.
* Even in the case of ONNX some layers are not supported by the new engine, such as all quantized layers (including DequantizeLinear, QuantizeLinear, QLinearConv etc.), LSTM, GRU, .... It's planned, of course, to have full support for ONNX by OpenCV 5.0 gold release. When a loaded model contains unsupported layers, we switch to the old engine automatically (at ONNX parsing time, not at `forward()` time).
* Some layers , e.g. Expat, are only partially supported by the new engine. In the case of unsupported flavours it switches to the old engine automatically (at ONNX parsing time, not at `forward()` time).
* 'Concat' graph optimization is disabled. The optimization eliminates Concat layer and instead makes the layers that generate tensors to be concatenated to write the outputs to the final destination. Of course, it's only possible when `axis=0` or `axis=N=1`. The optimization is not compatible with dynamic shapes since we need to know in advance where to store the tensors. Because some of the layer implementations have been modified to become more compatible with the new engine, the feature appears to be broken even when the old engine is used.
* Some `dnn::Net` API is not available with the new engine. Also, shape inference may return false if some of the output or intermediate tensors' shapes cannot be inferred without running the model. Probably this can be fixed by a dummy run of the model with zero inputs.
* Some overloads of `dnn::Net::getFLOPs()` and `dnn::Net::getMemoryConsumption()` are not exposed any longer in wrapper generators; but the most useful overloads are exposed (and checked by Java tests).
* [in progress] A few Einsum tests related to empty shapes have been disabled due to crashes in the tests and in Einsum implementations. The code and the tests need to be repaired.
* OpenCL implementation of Deconvolution is disabled. It's very bad and very slow anyway; need to be completely revised.
* Deconvolution3D test is now skipped, because it was only supported by CUDA and OpenVINO backends, both of which are not supported by the new engine.
* Some tests, such as FastNeuralStyle, checked that the in the case of CUDA backend there is no fallback to CPU. Currently all layers in the new engine are processed on CPU, so there are many fallbacks. The checks, therefore, have been temporarily disabled.
---
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] 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
* add Winograd FP16 implementation
* fixed dispatching of FP16 code paths in dnn; use dynamic dispatcher only when NEON_FP16 is enabled in the build and the feature is present in the host CPU at runtime
* fixed some warnings
* hopefully fixed winograd on x64 (and maybe other platforms)
---------
Co-authored-by: Vadim Pisarevsky <vadim.pisarevsky@gmail.com>
* attempt to add 0d/1d mat support to OpenCV
* revised the patch; now 1D mat is treated as 1xN 2D mat rather than Nx1.
* a step towards 'green' tests
* another little step towards 'green' tests
* calib test failures seem to be fixed now
* more fixes _core & _dnn
* another step towards green ci; even 0D mat's (a.k.a. scalars) are now partly supported!
* * fixed strange bug in aruco/charuco detector, not sure why it did not work
* also fixed a few remaining failures (hopefully) in dnn & core
* disabled failing GAPI tests - too complex to dig into this compiler pipeline
* hopefully fixed java tests
* trying to fix some more tests
* quick followup fix
* continue to fix test failures and warnings
* quick followup fix
* trying to fix some more tests
* partly fixed support for 0D/scalar UMat's
* use updated parseReduce() from upstream
* trying to fix the remaining test failures
* fixed [ch]aruco tests in Python
* still trying to fix tests
* revert "fix" in dnn's CUDA tensor
* trying to fix dnn+CUDA test failures
* fixed 1D umat creation
* hopefully fixed remaining cuda test failures
* removed training whitespaces
OCL_FP16 MatMul with large batch
* Workaround FP16 MatMul with large batch
* Fix OCL reinitialization
* Higher thresholds for INT8 quantization
* Try fix gemm_buffer_NT for half (columns)
* Fix GEMM by rows
* Add batch dimension to InnerProduct layer test
* Fix Test_ONNX_conformance.Layer_Test/test_basic_conv_with_padding
* Batch 16
* Replace all vload4
* Version suffix for MobileNetSSD_deploy Caffe model
DetectionOutput layer on OpenVINO without limitations #24069
### Pull Request Readiness Checklist
required for https://github.com/opencv/opencv/pull/23987
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
Switch to new OpenVINO API after 2022.1 release
* Pass Layer_Test_Convolution_DLDT.Accuracy/0 test
* Pass test Test_Caffe_layers.Softmax
* Failed 136 tests
* Fix Concat. Failed 120 tests
* Custom nGraph ops. 19 failed tests
* Set and get properties from Core
* Read model from buffer
* Change MaxPooling layer output names. Restore reshape
* Cosmetic changes
* Cosmetic changes
* Override getOutputsInfo
* Fixes for OpenVINO < 2022.1
* Async inference for 2021.4 and less
* Compile model with config
* Fix serialize for 2022.1
* Asynchronous inference with 2022.1
* Handle 1d outputs
* Work with model with dynamic output shape
* Fixes with 1d output for old API
* Control outputs by nGraph function for all OpenVINO versions
* Refer inputs in PrePostProcessor by indices
* Fix cycled dependency between InfEngineNgraphNode and InfEngineNgraphNet.
Add InferRequest callback only for async inference. Do not capture InferRequest object.
* Fix tests thresholds
* Fix HETERO:GPU,CPU plugin issues with unsupported layer
[GSoC] High Level API and Samples for Scene Text Detection and Recognition
* APIs and samples for scene text detection and recognition
* update APIs and tutorial for Text Detection and Recognition
* API updates:
(1) put decodeType into struct Voc
(2) optimize the post-processing of DB
* sample update:
(1) add transformation into scene_text_spotting.cpp
(2) modify text_detection.cpp with API update
* update tutorial
* simplify text recognition API
update tutorial
* update impl usage in recognize() and detect()
* dnn: refactoring public API of TextRecognitionModel/TextDetectionModel
* update provided models
update opencv.bib
* dnn: adjust text rectangle angle
* remove points ordering operation in model.cpp
* update gts of DB test in test_model.cpp
* dnn: ensure to keep text rectangle angle
- avoid 90/180 degree turns
* dnn(text): use quadrangle result in TextDetectionModel API
* dnn: update Text Detection API
(1) keep points' order consistent with (bl, tl, tr, br) in unclip
(2) update contourScore with boundingRect
Add option for NMS for boxes with different labels
* DetectionModel impl
* Add option for NMS for boxes with different labels
In the detect function in modules/dnn/include/opencv2/dnn/dnn.hpp, whose implementation can be found at modules/dnn/src/model.cpp, the Non Max Suppression (NMS) is applied only for objects of the same label. Thus, a flag
was added with the purpose to allow developers to choose if they want to keep the default implementation or wether they would like NMS to be applied to all the boxes, regardless of label.
The flag is called nmsDifferentLabels, and is given a default value of false, which applies the current default implementation, thus allowing existing projects to update opencv without disruption
Solves issue opencv#18832
* Change return type of set & Add default constr
* Add assertions due to default constructor
* enable tests for DNN_TARGET_CUDA_FP16
* disable deconvolution tests
* disable shortcut tests
* fix typos and some minor changes
* dnn(test): skip CUDA FP16 test too (run_pool_max)
Abhishek Gola
Vadim Pisarevsky
Maksim Shabunin
Alexander Smorkalov
Wanli
Alexander Alekhin
zihaomu
Dmitry Kurtaev
rogday
JIANG Yichen
Wenqing Zhang
Daniel Cauchi
YashasSamaga
Liubov Batanina
berak
Alexander Alekhin
Diego
Dmitry Kurtaev