OpenCV 4.3.0 release, Intel open source computer vision library

OpenCV 4.3.0 was released. OpenCV is Intel open source computer vision library that implements many common algorithms of image processing and computer vision.

This version update highlights include:

DNN module

• Improved layer / activation / support more models:

  • ONNX: LSTM, Broadcasting, algebraic constants, having a plurality of inputs Slice

  • DarkNet: packet convolution, S-type, swish, scale_channels

  • MobileNet SSD v3-:  # 16760

• The new samples / demos:

  • Clothes parts segmentation 和 CP-VTON

  • DaSiamRPN tracker

• Intel® inference engine backend (  OpenVINO ™  ):

  • By nGraph OpenVINO API adds support for custom layers  # 16628

  • OpenVINO API is used by default nGraph  # 16746

• Many repair and optimize CUDA backend

• OPEN AI LAB team used its Tengine library OpenCV DNN submitted may accelerate in the ARM patch:

• 

G-API module

• TBD

• Performance Improvements

  • IPP-ICV CPU optimization library has been updated to version 2020.0.0

  • SIMD internal integralfunction: , resize, (opencv_contrib) RLOF achieved  # 2476

Update Description: https://github.com/opencv/opencv/wiki/ChangeLog#version430