AILab-CVC / YOLO-World

[CVPR 2024] Real-Time Open-Vocabulary Object Detection
https://www.yoloworld.cc
GNU General Public License v3.0
4.64k stars 449 forks source link

Hugging face Demo #443

Open etemical opened 3 months ago

etemical commented 3 months ago

Why do the same prompt texts have results when used in the hugging face demo, but not in the yolo world program, Which configuration file and weights are used in the hugging face demo。

kaixin-bai commented 3 months ago

i have the same problem, i checked their huggingface code and find that they use the config file and checkpoint below:

configs/pretrain/yolo_world_xl_t2i_bn_2e-4_100e_4x8gpus_obj365v1_goldg_train_lvis_minival.py 

weights/yolo_world_v2_xl_obj365v1_goldg_cc3mlite_pretrain.pth

but even using the same config file and weight, i cannot get the same result as in the online demo

FantasticZihao commented 1 week ago

我遇到了同样的问题,我检查了他们的 huggingface 代码,发现他们使用了下面的配置文件和检查点:

configs/pretrain/yolo_world_xl_t2i_bn_2e-4_100e_4x8gpus_obj365v1_goldg_train_lvis_minival.py 

weights/yolo_world_v2_xl_obj365v1_goldg_cc3mlite_pretrain.pth

但是即使使用相同的配置文件和权重,我也无法获得与在线演示中相同的结果

Have you figured it out? I met the same question, but i use the package of ultralytics. I don't find the weights file you mentioned in ultralytics.