SPASE
A variant of the region-aware channel attention module based on superpixels, which uses the SLIC algorithm to generate adaptive region partitions.
Timeline (last 90 days)
The paper proposes four region-aware channel recalibration modules—BASE, SPASE, MAPSE and PLEASE—which define pooling regions using regular grids, SLIC superpixels, frozen network semantic maps and end-to-end learned soft assignments, respectively.
BASE and SPASE add almost no learnable parameters, making them suitable for scenarios sensitive to model size.
Experimental results show that the four modules surpass the original SE module and related spatial-channel attention variants on most benchmarks, achieving the best results in the majority of cases.
All Facts (3)
The paper proposes four region-aware channel recalibration modules—BASE, SPASE, MAPSE and PLEASE—which define pooling regions using regular grids, SLIC superpixels, frozen network semantic maps and end-to-end learned soft assignments, respectively.
50%UnverifiedBASE and SPASE add almost no learnable parameters, making them suitable for scenarios sensitive to model size.
50%UnverifiedExperimental results show that the four modules surpass the original SE module and related spatial-channel attention variants on most benchmarks, achieving the best results in the majority of cases.
50%