Boosting-R-CNN icon indicating copy to clipboard operation
Boosting-R-CNN copied to clipboard

[Neurocomputing 2023] An official implementation of Boosting R-CNN: Reweighting R-CNN Samples by RPN's Error for Underwater Object Detection

Results 17 Boosting-R-CNN issues
Sort by recently updated
recently updated
newest added

How to solve the metric Recall is not supported problem?

I have the following issue when loading checkpoint ![image](https://user-images.githubusercontent.com/81721256/196380796-7304a771-8a33-426b-9458-8317eed230fe.png) my environment: ![image](https://user-images.githubusercontent.com/81721256/196383695-4db51539-35f4-4ace-9fe8-c1e01e454ad4.png) this is my commend: python tools/test.py configs/boosting_rcnn/boosting_rcnn_r50_pafpn_1x_utdac.py ./ckpts/boosting_rcnn_r50_pafpn_1x_utdac.pth And I download the checkpoint is: ![image](https://user-images.githubusercontent.com/81721256/196384503-cb068bf1-e394-439c-b830-635908978a8e.png)

Does it look like there's no code module you changed? Like atss_rpn_head. Py

我的cuda版本是11.7,mmcv1.4版本太低,会冲突,但是安装高版本2.0.0的mmcv会报如下的错误,高版本集成方法的位置发生了变化,请问该怎么解决 ![image](https://github.com/user-attachments/assets/d403a3ad-01a2-4d3c-a090-ec28b2a29b2e)

您好,我看论文里面用到好几个数据集,有一张表说是泛化实验结果,但是不同数据集检测的目标类别都有所不同,例如UTDAC2020的数据集里面的标签和COCO的就不一样,跑COCO的时候是重新用COCO训练过吗?泛化不是直接用原本训练的模型在新数据集上推理预测,然后得到评估结果吗?