Small Bird Detection using YOLOv7 with Test-Time Augmentation
Kosuke Shigematsu
TL;DR
This work tackles the challenge of detecting small birds by applying YOLOv7 with test-time augmentation. The authors propose a concise three-part strategy: increase input resolution, perform multiscale and flipped-image inferences, and fuse results with weighted boxes fusion to improve precision for small objects. Experimental results show substantial gains over baseline, achieving a Development-category top score with a public AP of 0.732 and a private AP of 27.2 at IoU=0.5, demonstrating the effectiveness of TTA and WBF for small-object bird detection. The approach is lightweight and has potential impact on wildlife monitoring and agricultural pest management by improving reliable detection of small birds in challenging settings.
Abstract
In this paper, we propose a method specifically aimed at improving small bird detection for the Small Object Detection Challenge for Spotting Birds 2023. Utilizing YOLOv7 model with test-time augmentation, our approach involves increasing the input resolution, incorporating multiscale inference, considering flipped images during the inference process, and employing weighted boxes fusion to merge detection results. We rigorously explore the impact of each technique on detection performance. Experimental results demonstrate significant improvements in detection accuracy. Our method achieved a top score in the Development category, with a public AP of 0.732 and a private AP of 27.2, both at IoU=0.5.