eca49caee9
Use SAM2VideoPredictor.from_pretrained() instead of the checkpoint-based build_sam2_video_predictor() which doesn't accept HuggingFace model IDs. Threshold out_mask_logits at 0.0 and squeeze shape before converting to binary PNG. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>