The tag name is always person

[3083156185]

When I do video inference, I enter the cue text, e.g. aircraft, and save the resulting video, and it does detect the aircraft, but the label name is always person. i tried to print the inference result, and the mapped labels are always tensor([0]) . What is the cause of this, the detection is spot on but the tag name is always person.

python video_demo.py /root/autodl-tmp/YOLO-World/configs/pretrain/yolo_world_v2_l_vlpan_bn_2e-3_100e_4x8gpus_obj365v1_goldg_train_1280ft_lvis_minival.py /root/autodl-tmp/YOLO-World/pretrained_weights/yolo_world_v2_l_obj365v1_goldg_pretrain_1280ft-9babe3f6.pth /root/autodl-tmp/YOLO-World/source_data/9401.mp4 'airplane' --out /root/autodl-tmp/YOLO-World/source_data/5.mp4

[3083156185]

It didn't solve the problem.