Evaluation of Deep Models for Real-Time Small Object Detection

Phuoc Pham, Duy Nguyen, Tien Do, Thanh Duc Ngo, Duy-Dinh Le

Abstract: Real-time object detection is crucial for many applications. Approaches based on Deep Learning have achieved state-of-the-art performance on challenging datasets. Although several evaluations of the models have been conducted, there is no extensive evaluation with specific focuses on real-time small object detection. In this work, we present an in-depth evaluation of existing deep learning models in detecting small objects. We evaluate three state-of-the-art models including You Only Look Once (YOLO), Single Shot MultiBox Detector (SSD), and Faster R-CNN with related trade-off factors i.e. accuracy, execution time and resource constraints. Experiments were conducted on benchmark datasets and a newly generated dataset for small object detection. All analyses and findings are then presented.