Sampling Strategies for Efficient Training of Deep Learning Object Detection Algorithms

2025-05-23

AI Summary

To address the high annotation cost and low efficiency in training object detection models, this paper proposes two active sampling strategies grounded in the Lipschitz continuity assumption: (1) uniform sampling to ensure comprehensive coverage of the state space, and (2) frame-difference sampling to model temporal redundancy in videos and identify frames with high information gain. This work is the first to systematically incorporate Lipschitz continuity theory into detection data sampling design, establishing an interpretable and analytically tractable sampling framework; it also introduces the novel frame-difference sampling paradigm, substantially reducing reliance on dense manual annotations. Evaluated on COCO and a custom video object detection dataset, the proposed methods achieve up to 62–78% annotation reduction and ~45% fewer training iterations, while maintaining mAP degradation below 0.5%, demonstrating the feasibility of high-quality, efficient training under small-sample regimes.

Technology Category

• Computer Vision: Video Understanding & Activity Analysis
• Machine Learning: Active Learning
• Search and Optimization: Sampling/Simulation-based 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
• Security and Privacy: Large-scale security measurements

Abstract

Two sampling strategies are investigated to enhance efficiency in training a deep learning object detection model. These sampling strategies are employed under the assumption of Lipschitz continuity of deep learning models. The first strategy is uniform sampling which seeks to obtain samples evenly yet randomly through the state space of the object dynamics. The second strategy of frame difference sampling is developed to explore the temporal redundancy among successive frames in a video. Experiment result indicates that these proposed sampling strategies provide a dataset that yields good training performance while requiring relatively few manually labelled samples.

Problem

• Enhancing deep learning object detection training efficiency
• Exploring uniform and frame difference sampling strategies
• Reducing manual labeling with effective sampling techniques

Innovation

• Uniform sampling for even state space coverage
• Frame difference sampling reduces temporal redundancy
• Few labeled samples yield good training performance