search

dengxl0520 / MemSAM

[CVPR 2024 Oral] MemSAM: Taming Segment Anything Model for Echocardiography Video Segmentation.
MIT License
121 stars 11 forks source link

CAMUS dataset #5

Closed kimsekeun closed 3 months ago

kimsekeun commented 3 months ago

Dear author,

  1. CAMUS dataset only provides ED, ES annotations, could you please let me know where all frame annotated dataset is offered for evaluation?
  2. Could you please let me know about camus split indexes for training, val, and test set?
  3. For SonoSAM, the author of sonosam did not opened their code how did you evaluate on this method? and it needs point prompt for segmentation.
dengxl0520 commented 3 months ago

Thank you for your attention.

  1. Official CAMUS project has published full frame annotations.
  2. See here.
  3. We have only reproduced SonoSAM's method on our data, and the results presented are based on our reproduction only. The point prompts are randomly sampled from Ground Truth.