A Robust Approach for Small-Scale Object Detection From Aerial-View

In computer vision, object detection is an important task and gained significant progress but object detection in aerial images is still a challenging task for researchers. The small target size, low resolution, occlusion, attitude, and scale variations are the big concerns with aerial images that prevent many state-of-the-art object detectors to perform well. In our proposed approach, we have modified CenterNet and provided a comparison of experimentation results achieved using nine different CNN-based backbones i.e., resNet18, resNet34, resNet50, resNet101, resNet152, res2Net50, res2Net101, DLA34, and hourglass104. We found promising results using invariant of centerNet and hourglass104 as a backbone. We employed three challenging datasets to validate our approach i.e., VisDrone, Stanford, and AU-AIR. By keeping the standard mAP, we achieved 91.62, 75.62, and 34.85 validation results using AU-AIR, Stanford, and VisDrone datasets respectively. We have also compared the achieved mAP using IoU@0.5 and IoU@0.75 against different backbones. Our approach has achieved promising results as compared to the results in the latest research.