When W4A4 Breaks Camouflaged Object Detection: Token-Group Dual-Constraint Activation Quantization

TL;DR

This paper introduces COD-TDQ, a novel quantization method for Transformer-based camouflaged object detection that effectively handles token-range domination issues, improving performance on multiple benchmarks.

Contribution

The paper proposes a task-specific quantization approach with dual constraints to enhance low-bit inference for camouflaged object detection.

Findings

COD-TDQ outperforms state-of-the-art quantization methods by over 0.12 in Sα score.

It effectively suppresses cross-token range domination in 4-bit quantization.

The method improves detection accuracy across four benchmarks and two models.

Abstract

Camouflaged object detection (COD) segments objects that intentionally blend with the background, so predictions depend on subtle texture and boundary cues. COD is often needed under tight on-device memory and latency budgets, making low-bit inference highly desirable. However, COD is unusually hard to quantify aggressively. We study post-training W4A4 quantization of Transformer-based COD and find a task-specific cliff: heavy-tailed background tokens dominate a shared activation range, inflating the step size and pushing weak-but-structured boundary cues into the zero bin. This exposes a token-local bottleneck -- remove cross-token range domination and bound the zero-bin mass under 4-bit activations. To address this, we introduce COD-TDQ, a COD-aware Token-group Dual-constraint activation Quantization method. COD-TDQ addresses this token-local bottleneck with two coupled steps: Direct-Sum Token-Group (DSTG) assigns token-group scales to suppress cross-token range domination, and Dual-Constraint Range Projection (DCRP) projects each token-group clip range to keep the step-to-dispersion ratio and the zero-bin mass bounded. Across four COD benchmarks and two baseline models (CFRN and ESCNet), COD-TDQ consistently achieves an S{\alpha}score more than 0.12 higher than that of the state-of-the-art quantization method without retraining. The code will be released.