Zero-Shot Quantization for Object Detectors using Off-the-Shelf Generative Models

2026-06-30

AI Summary

This work addresses the significant performance degradation of object detection models under low-bit zero-shot quantization when original training data are unavailable. To tackle this challenge, the authors propose GoodQ, a novel framework that, for the first time, leverages off-the-shelf generative models for this task. GoodQ employs a three-stage strategy: it generates multi-instance images using information-dense prompts, mitigates class imbalance through intrinsic distribution-aware sampling, and suppresses pseudo-label noise via teacher-guided adaptive denoising. The method supports extremely low-bit quantization (W3A3) and achieves state-of-the-art performance under the W4A4 setting, substantially outperforming existing zero-shot quantization approaches.

Abstract

With an increasing number of Object Detection (OD) models being deployed on edge devices, Zero-Shot Quantization for OD (ZSQ-OD) aims to quantize these models when access to the original training data is prohibited. Existing research on Zero-Shot Quantization-Aware Training (QAT) for OD synthesizes training sets through noise optimization. However, this approach struggles to maintain performance in low-bit regions. In this paper, we introduce GoodQ (Generative off-the-shelf models for object detector Quantization), a QAT pipeline that utilizes off-the-shelf generative models to construct a training set. We first identify three challenges that arise when introducing a generative model to the ZSQ-OD task: 1) each image contains dense information with multiple instances, 2) the class-wise distribution in the original dataset is imbalanced, and 3) the pseudo-labels assigned to the generated images can potentially act as noisy signals during QAT. GoodQ addresses these challenges by 1) introducing an Information-Dense Prompting strategy to generate multi-instance images, 2) applying Intrinsic Distribution-Aware Selection to match the pretrained class distribution, and 3) employing Teacher-guided Adaptive Noise Reduction to mitigate noise arising from the QAT process. Our framework achieves state-of-the-art performance in low-bit ZSQ (W4A4) and extends quantization to extreme bit-widths (W3A3). Furthermore, we conduct an extensive analysis to uncover the underlying factors contributing to the efficacy of GoodQ.

Problem

• Zero-Shot Quantization
• Object Detection
• Quantization-Aware Training
• Low-bit Quantization
• Edge Devices

Innovation

• Zero-Shot Quantization
• Generative Models
• Object Detection
• Quantization-Aware Training
• Low-bit Quantization