shishaoyan

x_r总是负值，算法只能做亮度增强，无法使过曝区域变暗

1

x_r总是负值，根据model.py中增强代码，算法只能做亮度增强，无法使过曝区域变暗，这样看来论文中exp损失没有起到作用，是这样吗，且实验过程中发现图像增强后有些图像会变模糊请问是什么原因，

i want make image exposure,can it ?

about time consuming when training and test with query

2

I add p2 layer to yolov5m ,and modify p2 predict with query according to querydet， input image size is 256*256. Different from conclusion in paper, time consuming with query is...

Hi, we have recently updated a new VisDrone training config (gradient clip added), which can avoid NaN now.

2

Hi, we have recently updated a new VisDrone training config (gradient clip added), which can avoid NaN now. _Originally posted by @ChenhongyiYang in https://github.com/ChenhongyiYang/QueryDet-PyTorch/issues/40#issuecomment-1433519429_ I also meet this problem ,have...

I get miou 81.51 with using official setting in training cityscapes datasets , while official peper miou is 83.12, but I dont known why ,can you give me some advice?

hi, i am intereted in your work, and i reproduce paper's results with official parameters , but fid and colorfulness are both worse than paper, can you give me some...