The architecture of backbone
bigsquirrel18 opened this issue · comments
Hi, I printed architecture of the model, and found that the detection part is actually the same as that in yolov7.yaml rather than yolov7-e6e.yaml. And E-ELAN structure mentioned in the paper isn't the structure used in yolov7.yaml, but the one used in yolov7-e6e.yaml.
Btw, I'm also confused how to generate corresponding segmentation groundtruth if MixUp augmentation is performed in data processing. Since two images are mixed into one, and a pixel can belongs to both of the images. So how to decide which class does a pixel belong to?
Thank you very much for your attention to our work!
As this work is under study, these specific details are not convenient for discussion now. Please wait for specific follow-up progress and whether the company will release training codes.