CASIA-IVA-Lab / FastSAM

Fast Segment Anything
GNU Affero General Public License v3.0
7.1k stars 667 forks source link

object proposal cocoeval - how did you get your SAM result? #221

Open Winnie-Gong opened 2 months ago

Winnie-Gong commented 2 months ago

Amazing work, I'm trying to reproduce some results. How did you produce your /home/zhaoxu/yolo_sa/sam_h_20_coco_17val.json results? For cocoeval it requires a dict with keys: image_id, category_id,bbox, score I assume bbox and score are direct output of SAM (score being stability_score), and image_id would be the image ID of input image. How do you obtain category_id? What were the exact parameters when you set upSamAutomaticMaskGenerator? If you could release the code or even snippets on how you produced those results it would be amazing.

Thank you!

cfl-chenfangliang commented 2 months ago

you can call the function: FastSAM -> fastsam -> prompt#_format_results

Armadillooutfitter commented 2 months ago

it was on my computer