search

ArtyZe / yolov3_lite

yolov3 model compress and acceleration (quantization, sparse)， c++ version
37 stars 9 forks source link

Is this implementation for YOLOV2 or YOLOV3? #6

Open sicarioakki opened 5 years ago

ArtyZe commented 5 years ago

@sicarioakki up to now, schade not. some new layers in v3 in not realized in v2, but most changes are in convolutional_layer.c and parser.c, you can transplantation them to v3, later maybe I'll do it when I'm free ：）

sicarioakki commented 5 years ago

Okay! This is essentially for YOLOV2 then.. Thanks man!

abhigoku10 commented 5 years ago

@ArtyZe so you have incroprated the changes of pruning in convolutional_layer.c and parser.c right ?

katagoni commented 5 years ago

@sicarioakki up to now, schade not. some new layers in v3 in not realized in v2, but most changes are in convolutional_layer.c and parser.c, you can transplantation them to v3, later maybe I'll do it when I'm free ：）

Yolo3 weight please send me asd739@daum.net

ArtyZe commented 5 years ago

@katagoni This repo is now support v3, you can train your train sets like the original yolov3, the only thing you need to do is pick which accelerate method and open it in makefile, I have no idea what weight you need?

katagoni commented 5 years ago

@katagoni This repo is now support v3, you can train your train sets like the original yolov3, the only thing you need to do is pick which accelerate method and open it in makefile, I have no idea what weight you need?

How to train yolo3 i need recently strongest my favor

ArtyZe commented 5 years ago

@katagoni

  1. download my repo and open MASK, GPU, PRUNE，MULTI_CORE in makefile
  2. start to train just like the original yolov3 :) If you still don't know how to train, I can upload a txt file to explain how to use it (actually you should see the readme first)