data.md

May 26, 2024 · View on GitHub

Data	Description	Download
Visual Genome	`ln -s VG/images data/vg/images`	Official
MSCOCO 2014	`ln -s coco2014/images data/refcoco/images`	Official
MSCOCO 2017	`ln -s coco2017/val2017 data/coco2017/images`	Official
EVA CLIP text encoder	`mv <your_path>/ckpts/eva_clip_psz14.pt ckpts/`	Official
Converted annotations	`unzip data.zip`	OneDrive
Meteor package	`unzip meteor.zip -d dynrefer/common/evaluation/`	OneDrive
Pre-trained DynRefer weights and logs (Optional)	`mv <your_path>/ckpts/* ckpts/`	OneDrive

To train and evaluate DynRefer, download the files in the table and arrange the files according to the file tree below.

    |--DynRefer/
      |--data/
        |--vg/
           |--dynrefer/
           |--images/
              |--1000.jpg
              |--1001.jpg
              ...
        |--refcoco
           |--dynrefer/
           |--images/
              |--COCO_train2014_000000000009.jpg
              |--COCO_train2014_000000000025.jpg
              ...
        |--coco2017
           |--dynrefer/
           |--images/
              |--000000116207.jpg
              |--000000116208.jpg
              ...
      |--ckpts/
         |--eva_clip_psz14.pt
         |--vg1.2_5e.pth
         |--refcocog_ft.pth
      |--configs/
      |--dynrefer/
      |--docs/
      |--scripts/
      |--train.py
      |--eval.py