Evaluation Dataset Download

[StupidBuluchacha]

Hi ObjEmbed Team，

Can you release the script for downloading all the evaluation datasets? Or the official links for each dataset?

[fushh]

COCO: https://cocodataset.org/#home
LVIS: https://www.lvisdataset.org/
COCO-O: https://github.com/alibaba/easyrobust/tree/main/benchmarks/coco_o
odinw13: https://huggingface.co/GLIPModel/GLIP/tree/main/odinw_35
fg-ovd: https://github.com/lorebianchi98/FG-OVD/tree/main/benchmarks
d3: https://github.com/shikras/d-cube?tab=readme-ov-file#download
refcoco/+/g: https://huggingface.co/datasets/fushh7/eval_refcoco
sorce_1k: https://huggingface.co/datasets/lcxrocks/sorce-1k
reircoco: https://huggingface.co/datasets/haoxiangzhao/REIRCOCO
ilias: https://huggingface.co/datasets/vrg-prague/ilias
sharegpt-4v: https://github.com/ShareGPT4Omni/ShareGPT4V/blob/master/docs/Data.md
dci: https://github.com/facebookresearch/DCI
coco_caption_2017: https://cocodataset.org/#home
flickr30k: https://huggingface.co/datasets/nlphuji/flickr30k
coco_cn: We have provided.
flickr30k_cn: https://github.com/li-xirong/cross-lingual-cap

[StupidBuluchacha]

COCO: https://cocodataset.org/#home LVIS: https://www.lvisdataset.org/ COCO-O: https://github.com/alibaba/easyrobust/tree/main/benchmarks/coco_o odinw13: https://huggingface.co/GLIPModel/GLIP/tree/main/odinw_35 fg-ovd: https://github.com/lorebianchi98/FG-OVD/tree/main/benchmarks d3: https://github.com/shikras/d-cube?tab=readme-ov-file#download refcoco/+/g: https://huggingface.co/datasets/fushh7/eval_refcoco sorce_1k: https://huggingface.co/datasets/lcxrocks/sorce-1k reircoco: https://huggingface.co/datasets/haoxiangzhao/REIRCOCO ilias: https://huggingface.co/datasets/vrg-prague/ilias sharegpt-4v: https://github.com/ShareGPT4Omni/ShareGPT4V/blob/master/docs/Data.md dci: https://github.com/facebookresearch/DCI coco_caption_2017: https://cocodataset.org/#home flickr30k: https://huggingface.co/datasets/nlphuji/flickr30k coco_cn: We have provided. flickr30k_cn: https://github.com/li-xirong/cross-lingual-cap

Can we know how to configure this 'ann_path' for each evaluation dataset? Take sorce-1k as an example, how can we generate this file?

ObjEmbed/eval_retrieval/eval.py

Line 21 in ce99d33

'ann_path': 'datasets/sorce-1k/dataset.jsonl',

[fushh]

We have uploaded our processed files.