Skip to content

KooSung/megvii-traffic-sign-det

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

整体方案

  1. Faster R-CNN + ResNet-50
  2. 切图训练,整图推理

安装依赖

主要依赖版本:

  • python 3.8
  • megengie 1.5.0
pip install -r requirements.txt

预训练模型

wget -P ckpt/ https://data.megengine.org.cn/models/weights/faster_rcnn_res50_coco_3x_800size_40dot1_8682ff1a.pkl

数据预处理

训练集以800x800的尺寸切图,重叠率为20%,舍弃IoU小于 50%的bbox。

cd detection
# 训练集切图
python tools/crop_train.py
# 转换为coco格式
python tools/coco_train.py

训练

# 将sh文件里面的export PYTHONPATH=megvii-traffic-sign-det/detection:$PYTHONPATH修改成自己的路径
sh scrips/train.sh

推理

权重文件 https://drive.google.com/file/d/1j2MVCafoGgYqHzrFoMbGB_rmbigFc6VM/view?usp=sharing

下载后放到detection/workdirs/log-of-fr_r50_2x_bs_crop_800x800_min05_over02_freeze-0_recrop

# 将sh文件里面的export PYTHONPATH=megvii-traffic-sign-det/detection:$PYTHONPATH修改成自己的路径
sh scrips/infer.sh

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published