JarvisKevin

Results 5 comments of JarvisKevin

inconsistency with the original paper

I also found some difference between paper and code.

What is different in P5 model and P6 model?

And I found hyp.scratch.p6.yaml is the same as hyp.scratch.p5.yaml

High Resolution Images

maybe you can train on 256*256 image and inference in high resolution image.

[FEATURE] Vision Transformer Feature Extraction

The same problem present in other models, such as 'mobilenetv3_large_100'. The global_pool operator is embedded in the 'forward_features' function which is inconsistent with the tutorial in [https://rwightman.github.io/pytorch-image-models/feature_extraction/] @rwightman

loss_cls: 0.0000, loss_bbox: 0.0000 when training YOLOX on my own dataset

同！跑coco2017都不行