Request for Manual Review of Rejected SAM 3 Access — Academic Wood Ear Mushroom Instance Segmentation
Dear Meta SAM 3 Team,
I am writing to respectfully request a manual review or reset of my previously rejected access request for the facebook/sam3 model checkpoints.
I am a master's student working on a non-commercial academic project involving instance segmentation of wood ear mushrooms (Auricularia) in cultivation images.
I have already manually refined approximately 200 annotated images and recently processed another batch of approximately 80 images using SAM ViT-B.
However, the existing segmentation workflow still produces duplicate masks, fragmented instances, false positives on cultivation substrates, and difficulties separating overlapping mushroom instances.
I am particularly interested in evaluating SAM 3's text-prompted and exemplar-based segmentation capabilities to determine whether they can improve instance-level annotation quality and reduce the amount of manual correction required.
My intended use is strictly local, non-commercial academic experimentation. I will comply with the applicable license and usage requirements. I will not redistribute model checkpoints or provide unauthorized access to the model.
If my original application lacked sufficient detail, I apologize for the incomplete explanation and would appreciate an opportunity to clarify my research objectives.
Could you please reconsider my access request or reset its rejected status so that I may apply again?
Hugging Face username: SHAUN WANG
Academic affiliation:Harbin University of Commerce
Thank you for your time and consideration.
Best regards