OpenCV 4.4.0 is released. OpenCV is an Intel open source computer vision library that implements many common algorithms in image processing and computer vision.

　　Version update content:

The SIFT (Scale-Invariant Feature Transform) algorithm has been moved to the main repository (SIFT's patent has expired)

DNN module

Improved layers/activations/support for more models:

Latest Yolo v4 detector: #17148. Disabling per-layer NMS for [yolo] layers (Yolo v3 and Yolo v4) is incorrect - use cv::dnn::NMSBoxes in all detections.

ONNX: Add support for Resnet_backbone (Torchvision) #16887

EfficientDet model support: #17384

New samples/demos:

Add text recognition example: C++/Python

FlowNet2 optical flow: #16575

Intel® Inference Engine Backend (OpenVINO™):

Added support for OpenVINO 2020.3 LTS/2020.4 versions

Support for NN Builder API is planned to be removed in the next version

Many fixes and optimizations in the CUDA backend

G-API module

A new API for stateful kernels has been introduced in the OpenCV backend: GAPI_OCV_KERNEL_ST.

Introduced more video-oriented operations in G-API: goodFeaturesToTrack, buildOpticalFlowPyramid, calcOpicalFlowPyrLK.

More image processing kernels added: Laplacian and bilateral filters.

Fixed a potential crash in G-API's OpenCL backend.

　　Other important improvements

Obj-C/Swift bindings

BIMEF: A Bio-Inspired Multi-Exposure Fusion Framework for Low-light Image Enhancement

Text detection adds Stroke Width Transform algorithm