BOLIN

Training speed is very slow

1

@RogerChern I train tridentNet_1x with resnet50 on 4 GPU (a machine with 8 GPU), and I need 2 days. Especially, when others use other left GPUs in my machines, the...

Questions about shot_num

4

I have three questions about the shot_num setting. (1) In json_dataset.py，the parameter shot_num=10 (line 180+), and this means 10-shot setting? (2)in loader.py, I found the using 2-way training strategy, we...

In TFA/FSRW, FSOD fine-tune on both base and novel classes, using the same training examples provided by FSRW. And G-FSOD train for multiple runs on different random samples to obtain...

If there are no coco-related code and configs, we can not re-produce the method and can not use it as the baseline

Excuse me, can you provide the dataset you use because you mention that you only use a part of the full dataset. Thanks.