About [CVPR 2024] The official implementation of paper " Light the Night: A Multi-Condition Diffusion Framework for Unpaired Low-Light Enhancement in Autonomous Driving"
Apache License 2.0
55
stars
0
forks
source link
The result of 3D perception using the BEVDepth method #10
Thank you for your work! But I have some confusion.
As described in the paper, I used the only-front camera images of 616 daytime scenes and 15 nighttime scenes in nuScenes as the training set and the test set, respectively, with resnet-50 as the backbone for 24 epochs of training. Then the result I got was quite different from the paper
Could you explain the training configuration and the dataset you use in detail?
Thank you for your work! But I have some confusion. As described in the paper, I used the only-front camera images of 616 daytime scenes and 15 nighttime scenes in nuScenes as the training set and the test set, respectively, with resnet-50 as the backbone for 24 epochs of training. Then the result I got was quite different from the paper Could you explain the training configuration and the dataset you use in detail?