Skip to content

mattlovell/YoloX-Nano-Custom-Dataset-Training

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 

Repository files navigation

YoloX-Nano Custom Dataset Training

Yolox-Nano Custom Dataset training, based on https://github.com/Megvii-BaseDetection/YOLOX/

Have adapted the EXP example file in the above repo to allow for training a YoloX-Nano model using VOC dataset to train on custom classes.

Use the above repo, and add this file to yolox_voc_nano.py YOLOX/exps/example/yolox_voc/

Use this line once training project configured: !python tools/train.py -f exps/example/yolox_voc/yolox_voc_nano.py -d 1 -b 8 --fp16 -o -c /content/YOLOX/yolox_nano.pth

Tested on Google Colab Pro+

Total of 3200 images used to train 2 classes took approximately 5 hours and achieved best AP of 81.62

map_5095: 0.8161757313333597 map_50: 0.9539941110554803

Final model can be deloyed for use on mobile for close to real time object detection after ONNX conversion

Colab Notebook to follow...

About

Yolox-Nano Custom Dataset training, based on https://github.com/Megvii-BaseDetection/YOLOX/

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages