Author:Code-keys
Introduction
This project : YOLO-in-Darknet 🚀 is a project based on any YOLO-version
- The master branch is updating all version cfg files
- The Applications branch id built for real-usefully Deploy
- The v1.0-for-beginner branch is for studing Darknet framework and add special layer, such as add-SiLU
- The v2.0-for-torch-prune branch is aim to apply prune on nets
- The v3.0-for-tensorrt-C+ branch is a tensorrt-auto-transfer for *.cfg
- Aim to find a simple way to use all cnn-network by darknet.
- A implementation of official
- A implementation of scaled-yolov4 by darknet. including scaled-yolov4 p5、p6、p7、others.
- A implementation of Ultralytics-yolov5 using darknet. including s m l x.
- version 2 is Done :
- Merged two head-convs of the CSP to speed by group_id.
- version 4 is Done :
- Replace the Focus() by conv with stride=2 and groups=3.
- my yolov5-tiny may be better than yolov4-tiny: [paper url] .
- version 6 is Done :
- A implementation of Meituan-yolov6 using darknet. only YOLOv6s-3.0.
- Model demo for Deploy mode (Conv Bn fusioned )
- Exchange Convolution to VGGRepBlock for traning, and fusion layer for Deploy.
- yolov7x is Done ;
- yolov7 is Done ;
- yolov7-d6 is Done ;
- yolov8s is Done;
- yolov8n is Done ;
- yolov8m is Done ;
- yolov8l is Done ;
- yolov8x is Done ;
- SAM attention is Done ;
- CEM (SE) atttention is Done ;
darknet with SiLU implemented can seen at My another git as ' Darknet-plus '
yolo~ v3 serial
scaled-yolov4 serial
yolov4-tiny + SPP
yolov5 Ultralytics
yolov6 Meituan
yolov7 Official
yolov8 Official
special
thanks
- copyright©Code_keys