SegDis: Foreground Segmentation Followed by Type Discrimination for Salient and Camouflaged Object Detection
Organizations: Macau University of Science and Technology
Abstract
Salient and camouflaged object detection (SOD and COD) both recover foreground support, yet most joint methods output only a binary mask and therefore cannot distinguish the type. USC12K distinguishes the two types, but its baseline USCNet still binds foreground existence to the type within a single normalization, so the failure source cannot be attributed. Our cross-task study shows that foreground localization transfers between SOD and COD, whereas type discrimination does not: a COD-only model recovers foreground on SOD images but its type judgement follows the training distribution rather than the foreground-background relation, so the two decisions require separate mechanisms. SegDis factorizes the output into an explicit foreground probability and a conditional type probability given foreground: FLR provides high-quality task-agnostic support, and RTA refines the SOD/COD assignment from the foreground-background relation without a task identifier. Because support is explicit, each ground-truth foreground pixel is classified as correct-type, type-swapped, or missed, so foreground-recovery and type-assignment errors are diagnosed separately. With PVT-v2-B2, SegDis attains 68.98% FG-mIoU and 78.35% mIoU on USC12K at about one-eighth of USCNet's parameters; CSCS is 5.75.
Figures & tables
| Scene A | Scene B | Scene C | Overall scenes | |||||||||
| Method | Backbone | Total params (M) | IoU-S | IoU-C | IoU-S | IoU-C | IoU-S | IoU-C | FG-mIoU | mIoU | mAcc | CSCS |
| SOD-specific methods | ||||||||||||
| GateNet [ 26 ] | ResNeXt-101 | 68.32 | 54.26 | 66.85 | 35.03 | 65.08 | 44.17 | 54.63 | 68.27 | 78.07 | 11.30 | |
| F3Net [ 25 ] | ResNet-50 | 70.05 | 52.62 | 67.20 | 36.38 | 66.12 | 44.81 | 55.47 | 68.80 | 77.86 | 9.36 | |
| Auto-MSFNet [ 41 ] | ResNet-50 | 70.14 | 54.78 | 69.92 | 36.64 | 66.69 | 45.89 | 56.29 | 69.40 | 79.77 | 9.90 | |
| VST [ 27 ] | T2T-ViT-t-14 | 68.14 | 49.82 | 61.61 | 22.56 | 63.18 | 38.45 | 50.82 | 65.55 | 74.77 | 11.30 | |
| Benchmarks | bIoU | IoU-S | IoU-C | Type acc. | CSCS |
|---|---|---|---|---|---|
| SOD (5 sets) | 81.7 | 72.5 | 10.5 | 89.2 | 8.56 |
| COD (4 sets) | 70.9 | 3.5 | 70.8 | 95.0 | 6.42 |
Image Segmentation
Image segmentation assigns pixels or voxels to objects and regions, supporting measurement, scene understanding, and decisions across image domains.
See all 417 papers