Improved YOLOv7 for Small and Overlapping Objects Detection

Object detection is a crucial task in Deep Learning applications. However, detecting small, overlapping objects in complex scenes has always been challenging. To overcome this challenge, we propose a new algorithm based on YOLOv7, which incorporates a multi-path attention enhancement module in the neck part and a receptive field enhancement module on the feature layer responsible for forecasting small objects before the head. This allows all detection layers to focus more effectively on features that require attention and enhance receptive fields. Our experiments on the PASCAL VOC dataset demonstrate that our proposed approach significantly improves the detection of small and overlapping objects compared to YOLOv7.