yaml

[WHK1229]

Does sn2-yolov8n.yaml support version m？I added SNI and GSConvE to my V8 project and found that only n and s work, is it because n and s have the same depth and maximum number of channels? Changing m doesn't work.

[alanli1997]

Does sn2-yolov8n.yaml support version m？I added SNI and GSConvE to my V8 project and found that only n and s work, is it because n and s have the same depth and maximum number of channels? Changing m doesn't work.

Hi, we didn't test the m-size model of yolo so I'm not sure the GSConvE if will work well but I guess the SNI will be worked well. GSConvE focus on lightweight models, so it's efficiency maybe not obvious on m/l size models. By the way, you can try to add 2~3 k=3&s=1/k=1&s=1 convolutional layers in the SNI block on m/l size models for better features correcting.