Use coco_20k data?

Open longzeyilang opened this issue 1 year ago • 6 comments

Hi, what should I trianing RoMa with coco_20k data augmentation? like https://github.com/verlab/accelerated_features/blob/main/modules/dataset/augmentation.py https://github.com/verlab/accelerated_features/blob/main/modules/dataset/augmentation.py how should I revise?

Aug 13 '24 14:08 longzeyilang

I do not want use megadepth and scannet data, just like coco_20k data?

Aug 13 '24 14:08 longzeyilang

Basically you need to update the dataloader and loss. I used to have coco in DKM training. Wouldnt really recommend it though.

Aug 13 '24 14:08 Parskatt

why?

Aug 14 '24 00:08 longzeyilang

Because if you train dense matchers on homography they only do well on homography.

Aug 14 '24 01:08 Parskatt

ok , I want to train my own dataset, about 128*128 image size

Aug 14 '24 06:08 longzeyilang

According to you, I want use scannet data to crop 128*128 image size to train. and detection in my own dataset. is it correct?

Aug 15 '24 01:08 longzeyilang