fredzzhang / upt

[CVPR'22] Official PyTorch implementation for paper "Efficient Two-Stage Detection of Human–Object Interactions with a Novel Unary–Pairwise Transformer"
https://fredzzhang.com/unary-pairwise-transformers
BSD 3-Clause "New" or "Revised" License
144 stars 26 forks source link

How about training a DETR model on the VCOCO dataset #29

Closed ddwhzh closed 2 years ago

ddwhzh commented 2 years ago

Thank you for your excellent work. you have provided a tutorial on training DETR models on the HICO-DET dataset, could you tell us how you trained the DETR on the VCOCO dataset?

fredzzhang commented 2 years ago

Hi @ddwhzh,

The code for fine-tuning on V-COCO can be found in the same repository, under the branch vcoco, you do need to set up a softlink pointing to the V-COCO dataset though.

Cheers, Fred.

ddwhzh commented 2 years ago

Sorry for the possible misunderstanding due to the expression, I would like to know how to pre-train a detr pre-train model on vcoco instead of training upt with a pre-trained model.

fredzzhang commented 2 years ago

The vcoco branch is precisely for that purpose. If you look at the history of the branch, you'll find that commit 906fb89 added support for fine-tuning the object detector on V-COCO.

ddwhzh commented 2 years ago

THX, understood.