search

IDEA-Research / Grounded-Segment-Anything

Grounded SAM: Marrying Grounding DINO with Segment Anything & Stable Diffusion & Recognize Anything - Automatically Detect , Segment and Generate Anything
https://arxiv.org/abs/2401.14159
Apache License 2.0
15.16k stars 1.4k forks source link

为什么gardio_demo的效果比grounded_sam_demo的效果要差？ #73

Open huan085128 opened 1 year ago

huan085128 commented 1 year ago

都用的同样的与训练权重： config_file = 'GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py' ckpt_repo_id = "ShilongLiu/GroundingDINO" ckpt_filenmae = "groundingdino_swint_ogc.pth" sam_checkpoint='/home/Grounded-Segment-Anything/sam_vit_h_4b8939.pth'

gardio_demo的参数： Detection Prompt ： ’person' task type: det/seg/inpainting： seg Inpaint Prompt: 无 000022704 grounding_dino_output

Andy1621 commented 1 year ago

You can add NMS to remove overlapped boxes as follows: https://github.com/IDEA-Research/Grounded-Segment-Anything/blob/7c16d30a8cf5a876a88cbe28c0b91f205e625c71/automatic_label_demo.py#L274-L281