search

Megvii-BaseDetection / YOLOX

YOLOX is a high-performance anchor-free YOLO, exceeding yolov3~v5 with MegEngine, ONNX, TensorRT, ncnn, and OpenVINO supported. Documentation: https://yolox.readthedocs.io/
Apache License 2.0
9.43k stars 2.21k forks source link

Accuracy results #1017

Closed kolesnevg closed 2 years ago

kolesnevg commented 2 years ago

Hi, I'm trying to use YOLOX and I got one problem - I do not see Accuracy results at all About dataset - I converted my custom dataset to COCO format and use it

Then I trained model python tools/train.py -n yolox-tiny -d 1 -b 48 -o

After I run eval by command python tools/eval.py -n yolox-tiny -c /content/YOLOX/latest_ckpt.pth -b 64 -d 1 --conf 0.001

and got results

2021-12-22 13:42:53 | INFO | yolox.data.datasets.coco:45 - loading annotations into memory... 2021-12-22 13:42:53 | INFO | yolox.data.datasets.coco:45 - Done (t=0.00s) 2021-12-22 13:42:53 | INFO | pycocotools.coco:88 - creating index... 2021-12-22 13:42:53 | INFO | pycocotools.coco:88 - index created! 2021-12-22 13:42:57 | INFO | main:156 - loading checkpoint from /content/YOLOX/latest_ckpt.pth 2021-12-22 13:42:57 | INFO | main:160 - loaded checkpoint done. 100%|##########| 12/12 [00:10<00:00, 1.12it/s] 2021-12-22 13:43:08 | INFO | yolox.evaluators.coco_evaluator:171 - Evaluate in main process... 2021-12-22 13:43:08 | INFO | main:187 - Average forward time: 1.03 ms, Average NMS time: 0.17 ms, Average inference time: 1.20 ms

There is no accuracy :(

FateScript commented 2 years ago

The log show that your model detected nothing.

kolesnevg commented 2 years ago

Thank you, the problem was in my dataset annotations