few-shot-object-detection
few-shot-object-detection copied to clipboard
ucbdrive
The results of 3shot on VOC Split1
I find the results of 3shot are lower than that of the paper. The nAP50 is 38.3.
We expect there to be tiny differences between runs, so the numbers won't be exactly the same. You can try rerunning the training to see what the results are.
We expect there to be tiny differences between runs, so the numbers won't be exactly the same. You can try rerunning the training to see what the results are.
The paper result of split0 shot3 is 43.6,really? this is not a tiny difference~
The results could vary depending on your base model as well. We found the results to be fairly consistent with the same base model.
@thomasehuang I can get the similar split1 5shot results with the reported one, ~ 55% nAP50 with the base model you released. However, if I train the base model by myself, it will degrade to ~52%. Note that my base model indeed has a consistent base mAP with yours, e.g., ~81% for AP50. Could you advise? Thanks.
