复现过程中的问题

[xiaohui0225]

学长您好，我对超分辨率图像目标检测很感兴趣，您的工作非常棒！我在复现过程中有几个疑问想请教您一下。 1、UAVDT2COCO.py、VisDrone2COCO.py、UFP_VisDrone2COCO.py之间有什么关系吗，特别是后面两个。 2、UFP和MP-Det是分开训练的吧？UFP的训练集是原数据，而MP-Det的训练集是UFP的输出结果，这样理解对吗？ 2、在最终得到检测结果后，将检测结果还原到原超分辨率图像的code在哪呢？

[PuAnysh]

Sorry for replay so late:

1. To train the detector conveniently by mmdetection framework, UAVDT2COCO.py and VisDrone2COCO.py are to convert the original data set to coco format and facilitate the training of the mmdetection framework. UFP_ VisDrone2COCO. Py obtains the spliced image data set by a coarse detector.
2. Yes
3. https://github.com/PuAnysh/UFPMP-Det/blob/main/UFPMP-Det-Tools/eval_script/ufpmp_det_eval.py#L268

[jerryniu0624]

请问有具体的mmcv版本吗？我按照Readme下环境结果报错No module named 'mmcv.runner'

[peiqi999]

Sorry for replay so late:

1. To train the detector conveniently by mmdetection framework, UAVDT2COCO.py and VisDrone2COCO.py are to convert the original data set to coco format and facilitate the training of the mmdetection framework. UFP_ VisDrone2COCO. Py obtains the spliced image data set by a coarse detector.
2. Yes
3. https://github.com/PuAnysh/UFPMP-Det/blob/main/UFPMP-Det-Tools/eval_script/ufpmp_det_eval.py#L268

@PuAnysh I have encountered AttributeError: module 'mmdet. models. backbones' has no attribute' MPHead '. How can I resolve it