-
Dear author, how to fine-tune the retinanet model? I customized my own data set according to the train.json in the demo, and used python tools/train.py retinanet_ResNeXt101.py to train my own data set…
-
When i use focal loss, not mse_loss, the map will only 22%. However, the MAP will arrive 86% when using mse_loss. And I set hm_gauss = 4.
How to solve the problem, tks!
And did the code default onl…
-
Hello,
I am experiencing some odd behavior when trying to train on a custom dataset in COCO format. So far I have tried training faster-rcnn and retinanet models but the bbox loss is always zero. T…
-
I have 1 GPU,I run the train.py :
python train.py /media/david/fj/wx/SpatialFlow/SpatialFlow/configs/spatialflow/spatialflow_r50_fpn_20e_coco.py
I met a question as follows.I don't know how to s…
-
Hi @zylo117, thanks for the great work. However, I have a small question regarding your FocalLoss implementation.
Inferred from your [loss.py](https://github.com/zylo117/Yet-Another-EfficientDet-P…
-
I have 2 titan xp gpus ,and when I run reppoints detection train.py it occurs:
```
python tools/train.py workproject/gureppoints/reppoints_moment_r101_fpn_gn-neck+head_2x_coco.py --gpu-ids=1
…
-
Thanks for your error report and we appreciate it a lot.
**Checklist**
1. I have searched related issues but cannot get the expected help.
2. The bug has not been fixed in the latest version.
…
-
a subtask of #84, port a focal loss function into Monai
e.g.:
https://arxiv.org/abs/1708.02002
-
When we use Flair for Multi-Label Text Classification, we use the BCELoss. This works great but suffers badly in cases of class-imbalances in the data.
An alternative for the same could be FocalLoss …
-
Hi @zylo117 ,
I've seen that in [f833223](https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch/commit/f833223fa6687e0062924d4d649a80d9107622e4#diff-34fc60f6e9f3fced47a7a4b622abc0c0L128) you'…