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
9.44k stars 2.21k forks source link

COCO dataset train log #767

Open godbaz opened 3 years ago

godbaz commented 3 years ago

Thx for your great work. Can you share COCO dataset train log from start? YOLOX_nano, YOLOX_tiny, YOLOX_s, YOLOX_m, YOLOX_l, YOLOX_x。 Thx!

infinityglow commented 3 years ago

Here is my training log.

Screenshot 2021-10-11 at 10 49 40 AM
godbaz commented 3 years ago

Here is my training log. Screenshot 2021-10-11 at 10 49 40 AM

Thx for your reply. Can you share whole YOLOX-Tiny train log on coco dataset? I want to reproduce paper mAP@IoU=0.5:0.95=25.8 from training start.

infinityglow commented 3 years ago

Here is my training log. Screenshot 2021-10-11 at 10 49 40 AM

Thx for your reply. Can you share whole YOLOX-Tiny train log on coco dataset? I want to reproduce paper mAP@IoU=0.5:0.95=25.8 from training start.

I am training YOLOX-Small model on the GPU, it will show the result after several hours, and I used a custom dataset instead of COCO2017. I can share the whole train log with you after training.

Joker316701882 commented 3 years ago

@shl1993 You can get the training log for YOLOX-S here: https://github.com/Megvii-BaseDetection/YOLOX/issues/735#issuecomment-940607639

But noted that you need to change the decay_factor to 0.9999

godbaz commented 3 years ago

@shl1993 You can get the training log for YOLOX-S here: #735 (comment)

But noted that you need to change the decay_factor to 0.9999

Thx for your share! I'm reproduce YOLOX-Nano, can you share training log of YOLOX-Nano?

nemonameless commented 2 years ago

@Joker316701882 could you release all the training logs of yolox from nano to X？thx