Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

cocoapi does not handle empty results very well #24

Open
fu-7 opened this issue Nov 9, 2022 · 9 comments
Open

cocoapi does not handle empty results very well #24

fu-7 opened this issue Nov 9, 2022 · 9 comments

Comments

@fu-7
Copy link

fu-7 commented Nov 9, 2022

您好,在运行coco-full时遇到以下bug,请问一下您知道是怎么回事吗?
` File "/root/DenseTeacher/cvpods/cvpods/evaluation/coco_evaluation.py", line 152, in evaluate
self._eval_predictions(set(self._tasks))
│ │ │ └ ('bbox', 'segm')
│ │ └ <cvpods.evaluation.coco_evaluation.COCOEvaluator object at 0x7f3b800ad210>
│ └ <function COCOEvaluator._eval_predictions at 0x7f3caa137830>
└ <cvpods.evaluation.coco_evaluation.COCOEvaluator object at 0x7f3b800ad210>

File "/root/DenseTeacher/cvpods/cvpods/evaluation/coco_evaluation.py", line 238, in _eval_predictions
else None # cocoapi does not handle empty results very well`

@ZRandomize
Copy link
Collaborator

似乎是训练发散了,可以重跑一次试试

@fu-7
Copy link
Author

fu-7 commented Nov 25, 2022

请问你知道teacher模型与student模型精度相差这么大的原因吗?

@ZRandomize
Copy link
Collaborator

这两个性能gap由EMA和strong aug导致

@LiuXiaolong19920720
Copy link

在训练自定义数据集(class number = 1)的时候,遇到了同样的问题。而且保存下来的模型,所有图片结果也都为空,请问问题解决了吗?有没有什么建议? @ZRandomize @fu-7

@fu-7
Copy link
Author

fu-7 commented Dec 2, 2022

在训练自定义数据集(class number = 1)的时候,遇到了同样的问题。而且保存下来的模型,所有图片结果也都为空,请问问题解决了吗?有没有什么建议? @ZRandomize @fu-7
我的确重新训练一次就没有出现这个问题了,具体原因不清楚

@LiuXiaolong19920720
Copy link

确实有时不会出现这个问题,但是模型输出为空

@ZRandomize
Copy link
Collaborator

这个问题确实存在,而且和半监督似乎没有关系,因为空预测可以在burn in之前出现。

@LiuXiaolong19920720
Copy link

TRAINER.EMA.START_STEPS > TEST.EVAL_PERIOD的时候,会出现cocoapi does not handle empty results very well的问题。TEST.EVAL_PERIOD的实际数值会根据卡数变化,需要注意实际的TEST.EVAL_PERIOD数值。供后来者参考。

@BroenLin
Copy link

BroenLin commented Feb 12, 2023

在训练自定义数据集(class number = 1)的时候,遇到了同样的问题。而且保存下来的模型,所有图片结果也都为空,请问问题解决了吗?有没有什么建议? @ZRandomize @fu-7
我的确重新训练一次就没有出现这个问题了,具体原因不清楚

请问如何更改config中的num_classes。我已经对fcos.py进行了更改,但是训练的时候model config依旧显示num_classes=80,训练到model_iter_0200000.pth时,也因为category_id=56报错
config
fcos

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants