PTQ4RIS: Post-Training Quantization for Referring Image Segmentation

TL;DR

PTQ4RIS introduces a novel post-training quantization framework tailored for referring image segmentation, enabling efficient on-device inference without significant performance loss.

Contribution

This work is the first to develop a PTQ method specifically for RIS, addressing quantization challenges in visual and linguistic encoders with novel techniques.

Findings

Achieves superior performance across multiple benchmarks

Supports quantization from 8 to 4 bits with minimal accuracy loss

Demonstrates feasibility of PTQ for RIS applications

Abstract

Referring Image Segmentation (RIS), aims to segment the object referred by a given sentence in an image by understanding both visual and linguistic information. However, existing RIS methods tend to explore top-performance models, disregarding considerations for practical applications on resources-limited edge devices. This oversight poses a significant challenge for on-device RIS inference. To this end, we propose an effective and efficient post-training quantization framework termed PTQ4RIS. Specifically, we first conduct an in-depth analysis of the root causes of performance degradation in RIS model quantization and propose dual-region quantization (DRQ) and reorder-based outlier-retained quantization (RORQ) to address the quantization difficulties in visual and text encoders. Extensive experiments on three benchmarks with different bits settings (from 8 to 4 bits) demonstrates its superior performance. Importantly, we are the first PTQ method specifically designed for the RIS task, highlighting the feasibility of PTQ in RIS applications. Code and video are available at {https://github.com/gugu511yy/PTQ4RIS}.