Skip to content

Latest commit

 

History

History
37 lines (34 loc) · 853 Bytes

README.md

File metadata and controls

37 lines (34 loc) · 853 Bytes
Raw

pl-YOLO

Detection on pytorch lightning.

Step of training

1. Build the model parameters and dataset parameters in configs.

2. Run

  python train.py -c <path to the model yaml> -d <path to the data yaml>

Example

COCO2017 Dataset

Dataset directory

COCO2017  
└─annotations_trainval2017   
│   └─annotations   
│      │ instances_train2017.json   
│      │ instances_train2017.json   
└─train2017
└─val2017

Change dataset parameters

In configs/data/coco2017.yaml, change the directory, image size and batch size.

    dir: <path to your COCO2017>
    ...
    train_size: [640,640]
    val_size: [640,640]
    train_batch_size: 32
    val_batch_size: 32

Run YOLOX-s

  python train.py -d configs/data/coco2017.yaml -c configs/model/yolox/yolox_s.yaml