STK-YOLO

The VisDrone2019 dataset [34], released by theAISKYEYE team of the Machine Learning and DataMining Laboratory at Tianjin University, is a large\u0002scale benchmark specifically designed for UAV-based object detection tasks. The dataset comprises 10,209high-resolution static images, captured across 14 citiesin China, covering diverse environmental conditions, in\u0002cluding daytime and nighttime scenarios. The dataset isdivided into four subsets: 6,471 training images, 548validation images, 1,610 test images, and 1,580 challengeimages. Images were acquired using professional UAVphotography equipment, ensuring high quality and richdetails