AxisQ improves camouflaged object detection using depth-adaptive quantization
AxisQ: Depth-Adaptive Channel--Frequency Quantization for Camouflaged Object Detection
Computer Vision and Pattern Recognition
Summary
Detecting camouflaged objects is hard because their edges and textures are very faint, and simple ways to shrink models can ruin these clues. The authors designed AxisQ, a method that changes how finely different parts of the model’s features are stored depending on how deep they are in the network. For layers closer to the input, it focuses on channels, and for deeper layers, on local frequency details. This clever approach helps keep the important subtle information, making smaller models that can still find camouflaged things more accurately.
What this means in practice
- •For mobile app developers: Build lightweight camouflaged object detection models that run efficiently on mobile devices without losing accuracy in subtle regions.
- •For security system engineers: Create resource-efficient surveillance solutions that better detect hard-to-see objects by preserving boundary details in compressed models.
Authors
Chi Ruan, Tianqi Li, Xin He, Xue Geng, Ming-Ming Cheng, Yun Liu
Abstract
Camouflaged object detection (COD) depends on weak texture and boundary cues that can be destroyed by aggressive post-training quantization (PTQ). Existing PTQ methods mainly optimize quantizers or reconstruct network blocks, while using a single precision-allocation domain throughout hierarchical backbones. Instead, we introduce AxisQ, a depth-adaptive bit-allocation framework motivated by the observation that the useful allocation axis changes with feature depth. Under a fixed per-layer average activation-bit budget, AxisQ allocates precision in the channel domain for shallow features and in the local-frequency domain for deep features, using COD-aware sensitivity to better preserve subtle foreground regions and ambiguous boundaries. We also formulate a boundary-aware Hessian approximation for reconstruction by incorporating boundary-weighted gradient and error statistics. Experiments across multiple COD models and diverse benchmarks demonstrate improved accuracy over the strongest baselines at low bit widths, with relative weighted F-measure gains of 12.7% for CFRN and 33.3% for CamoFormer-P on CAMO under W2A2 quantization. These results support adapting the allocation domain to feature depth as a transferable quantization strategy for hierarchical COD models.