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
9.48k stars 2.22k forks source link

AP=0 and AR=0 in COCO datasets #492

Open Kangzf1996 opened 3 years ago

Kangzf1996 commented 3 years ago

During the training, AP is always 0. btw, I use the COCO datasets.

And I use this command, python tools/train.py -f exps/example/yolox_s.py -b 64 --fp16 -o

Average forward time: 1.82 ms, Average NMS time: 0.65 ms, Average inference time: 2.47 ms Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.001 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.001 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.002

Joker316701882 commented 3 years ago

Is this the result for 10th epoch?

kikyoyang commented 3 years ago

yes @Joker316701882

Joker316701882 commented 3 years ago

@kikyoyang Please check your data. We can get proper training results using the current code.

Kangzf1996 commented 3 years ago

@kikyoyang Please check your data. We can get proper training results using the current code.

I used the COCO datasets. The data is ok.

18789447455 commented 3 years ago

请问你解决这个问题了吗?我也遇到这个问题。

DoubleWJX commented 2 years ago

@kikyoyang I meet the same problem. I fix this by delete the cache data and rerun the experiments, which works for me!

cache data path: path_to_coco/img_resized_cache_train2017.array

please be aware of the following logs:

image

2catycm commented 2 years ago

@kikyoyang I meet the same problem. I fix this by delete the cache data and rerun the experiments, which works for me!

cache data path: path_to_coco/img_resized_cache_train2017.array

please be aware of the following logs:

image

thanks, your answer is exactly for what I encountered. what if after I deleted it, it still ask to cache again?