Pytorch -> ONNX -> ONNXRuntime/OpenCV/MNN/TensorRT/OpenVINO
As a computer vision engineer, how to better apply image algorithms to landing scenes is crucial. In practice, C++ can provide a faster reasoning speed and a more practical deployment platform; In addition, Python can provide more convenient simulation and processing.
- 2.4.0 NNAPI后端/CUDA后端支持量化模型
- ONNX Runtime v1.14.1
- OpenCV 4.7.0
- ultralytics/yolov5 v7.0 - YOLOv5 SOTA Realtime Instance Segmentation
[ERROR:0@2.663] global onnx_importer.cpp:1051 handleNode DNN/ONNX: ERROR during processing node with 2 inputs and 3 outputs: [Split]:(onnx_node!/model.24/Split) from domain='ai.onnx'
- global onnx_importer.cpp:1051 handleNode DNN/ONNX
- OPENCV部署ONNX模型报错 ERROR during processing node with 1 inputs and 1 outputs
In short, OpenCV 4.7.0 only supports ONNX models with fixed input sizes, and this issue will be resolved after the 5. X. X series
- zhujian - Initial work - zjykzj
- Open Neural Network Exchange
- pytorch/pytorch
- pytorch/vision
- alibaba/MNN
- microsoft/onnxruntime
- rockchip-linux/rknn-toolkit2
- libjpeg-turbo/libjpeg-turbo
- opencv/opencv
- opencv/opencv-python
- ermig1979/Simd
- nothings/stb
- gabime/spdlog
- facebookresearch/faiss
- NVIDIA/TensorRT
- NVIDIA Deep Learning TensorRT Documentation
Anyone's participation is welcome! Open an issue or submit PRs.
Small note:
- Git submission specifications should be complied with Conventional Commits
- If versioned, please conform to the Semantic Versioning 2.0.0 specification
- If editing the README, please conform to the standard-readme specification.
Apache License 2.0 © 2021 zjykzj