Accurate and Efficient Two-Stage Gun Detection in Video

2025-03-08

AI Summary

To address the challenges of detecting small-scale, appearance-varying, and sparsely annotated firearms in videos, this paper proposes a two-stage collaborative detection framework—“classify-then-localize.” It innovatively decouples video-level firearm existence classification from frame-level precise localization: a lightweight enhanced video classifier performs coarse-grained clip-level screening, while a jointly optimized YOLO-style detector achieves fine-grained spatial localization. Customized data augmentation strategies, loss functions, and evaluation protocols are further introduced to accommodate the unique characteristics of firearm detection. Evaluated on real-world video datasets, the method achieves a 12.6% improvement in mean Average Precision (mAP) and operates at 23 FPS, significantly outperforming both single-stage detectors and generic video models. The framework strikes a superior balance between detection accuracy and real-time inference capability.

Technology Category

• Computer Vision: Object Detection & Categorization
• Machine Learning: Hardware-aware ML
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Object detection in videos plays a crucial role in advancing applications such as public safety and anomaly detection. Existing methods have explored different techniques, including CNN, deep learning, and Transformers, for object detection and video classification. However, detecting tiny objects, e.g., guns, in videos remains challenging due to their small scale and varying appearances in complex scenes. Moreover, existing video analysis models for classification or detection often perform poorly in real-world gun detection scenarios due to limited labeled video datasets for training. Thus, developing efficient methods for effectively capturing tiny object features and designing models capable of accurate gun detection in real-world videos is imperative. To address these challenges, we make three original contributions in this paper. First, we conduct an empirical study of several existing video classification and object detection methods to identify guns in videos. Our extensive analysis shows that these methods may not accurately detect guns in videos. Second, we propose a novel two-stage gun detection method. In stage 1, we train an image-augmented model to effectively classify ``Gun'' videos. To make the detection more precise and efficient, stage 2 employs an object detection model to locate the exact region of the gun within video frames for videos classified as ``Gun'' by stage 1. Third, our experimental results demonstrate that the proposed domain-specific method achieves significant performance improvements and enhances efficiency compared with existing techniques. We also discuss challenges and future research directions in gun detection tasks in computer vision.

Problem

• Detecting tiny objects like guns in videos is challenging due to small scale and varying appearances.
• Existing models perform poorly in real-world gun detection due to limited labeled datasets.
• A two-stage method is proposed to improve accuracy and efficiency in gun detection.

Innovation

• Two-stage method for gun detection
• Image-augmented model for video classification
• Object detection model for precise gun localization