Task-Specific Zero-shot Quantization-Aware Training for Object Detection

2025-07-22

AI Summary

Traditional zero-shot quantization (ZSQ) for object detection suffers from severe performance degradation due to reliance on unlabeled, task-agnostic synthetic calibration data. This work proposes a task-specific zero-shot quantization-aware training framework: it implicitly reconstructs object location, scale, and category distributions directly from a pre-trained detector, generating task-aligned synthetic data with bounding boxes and class labels—without requiring real data or human priors. This enables joint optimization of quantization-aware training and knowledge distillation. Evaluated on MS-COCO and Pascal VOC, our method significantly outperforms existing ZSQ approaches, achieving state-of-the-art accuracy under zero-shot settings. The implementation is publicly available.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Object Detection & Categorization
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Quantization is a key technique to reduce network size and computational complexity by representing the network parameters with a lower precision. Traditional quantization methods rely on access to original training data, which is often restricted due to privacy concerns or security challenges. Zero-shot Quantization (ZSQ) addresses this by using synthetic data generated from pre-trained models, eliminating the need for real training data. Recently, ZSQ has been extended to object detection. However, existing methods use unlabeled task-agnostic synthetic images that lack the specific information required for object detection, leading to suboptimal performance. In this paper, we propose a novel task-specific ZSQ framework for object detection networks, which consists of two main stages. First, we introduce a bounding box and category sampling strategy to synthesize a task-specific calibration set from the pre-trained network, reconstructing object locations, sizes, and category distributions without any prior knowledge. Second, we integrate task-specific training into the knowledge distillation process to restore the performance of quantized detection networks. Extensive experiments conducted on the MS-COCO and Pascal VOC datasets demonstrate the efficiency and state-of-the-art performance of our method. Our code is publicly available at: https://github.com/DFQ-Dojo/dfq-toolkit .

Problem

• Improves zero-shot quantization for object detection networks
• Generates task-specific synthetic data for better quantization
• Enhances performance without needing original training data

Innovation

• Task-specific synthetic data for object detection
• Bounding box and category sampling strategy
• Task-specific knowledge distillation process