search

cuiziteng / Illumination-Adaptive-Transformer

[BMVC 2022] You Only Need 90K Parameters to Adapt Light: A Light Weight Transformer for Image Enhancement and Exposure Correction. SOTA for low light enhancement, 0.004 seconds try this for pre-processing.
Apache License 2.0
439 stars 43 forks source link

您好，请问IAT与yolov3联合训练的时候，loss函数中似乎没有IAT的损失函数 ，只有yolov3的 #72

Open lissongod opened 3 months ago

lissongod commented 3 months ago

您好，请问IAT与yolov3联合训练的时候，loss函数中似乎没有IAT的损失函数 ，只有yolov3的

cuiziteng commented 3 months ago

Yes, we directly add the perception loss function, because only the detection loss is enough, the object detection task is concentrate on detection performance rather than image quality.

In fact the follow-up work also do not add image reconstruction loss function, like FeatEnhancer: https://openaccess.thecvf.com/content/ICCV2023/papers/Hashmi_FeatEnHancer_Enhancing_Hierarchical_Features_for_Object_Detection_and_Beyond_Under_ICCV_2023_paper.pdf