Experimental effect on voc dataset

[hero-y]

Hi, your work has a good effect on the coco dataset. Do you have made experiments on the voc dataset? I use the same parameters on the voc dataset, but it has a lower ap.5,ap.75,ap.8 than origin retinanet, why is it? Thank you!

[libuyu]

Actually, the results on Pascal Voc have been listed on the poster of my work, and the AP of Focal Loss and GHM are 74.5 vs 74.8. And for training on voc, the warmup iterations should be large enough, like 2k. If not, either Focal Loss or GHM can not get a reasonable result.

[hero-y]

Where is the poster of your work? Can you give me a link? Thank you!

[libuyu]

I can share with you my slides for the oral presentation, it has more details than the poster. And the results on VOC are on page 10.
https://drive.google.com/open?id=1H3tfg2d3NdLPQ7HHtZSgecE7ybN7Z-uS

[hero-y]

Is the result AP.5? Do you have made experiments using both GHM-C and GHM-R on the voc dataset?My results show that origin retinanet is 0.793，0.560，0.176 about the ap.5,ap.75,ap.9, and the ghm using ghmc and ghmr is 0.779,0.553,0.207.