BED-SAM2: Boundary-Enhanced-Depth SAM2 via Monocular Geometric Priors
About
Building upon the SAM2 vision foundation model for downstream segmentation, this study introduces Boundary Enhanced Depth (BED)-SAM2. The SAM2 Hiera encoder architecture is modified to directly encode monocular depth information from RGB images, thereby providing geometric cues that enhance object boundary delineation and facilitate the extraction of camouflaged object shapes. BED-SAM2 demonstrates competitive state-of-the-art performance across multiple salient and camouflaged object detection tasks with as few as five training epochs.
Tyler Rust, Dara McNally, Kyle O'Donnell, Colin Kelly, Chandra Kambhamettu• 2026
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Salient Object Detection | DUTS (test) | M (MAE)0.017 | 357 | |
| Camouflaged Object Detection | NC4K (test) | Sm0.912 | 89 | |
| Salient Object Detection | ECSSD 1,000 images (test) | MAE0.017 | 68 | |
| Camouflaged Object Detection | CAMO 250 (test) | M (Mean Score)0.04 | 65 | |
| Salient Object Detection | HKU-IS 4,447 images (test) | MAE0.018 | 62 | |
| Salient Object Detection | PASCAL-S 850 images (test) | MAE0.041 | 61 | |
| Salient Object Detection | NJU2K | S-measure (Sα)92.5 | 46 | |
| Camouflaged Object Detection | CHAMELEON 76 (test) | Sm0.941 | 44 | |
| Salient Object Detection | DUT-OMRON 5168 images (test) | Sm89.8 | 29 | |
| Salient Object Detection | NLPR 300 | Sm Score94.2 | 23 |
Showing 10 of 13 rows