Modern PyTorch toolkit for the VisDrone aerial object detection dataset with production-ready training pipelines, real-time inference, and format converters. Features state-of-the-art models (Faster R-CNN, FCOS, RetinaNet), mixed precision training, rich progress tracking, and optimizations for small object detection in drone imagery.

Many yolov8 model are trained on the VisDrone dataset.

基于渐进式学习的轻量级无人机目标检测

object detection using YOLO11x on VisDrone dataset

Object detection format converter from VisDrone2019-DET to Yolo.

Attention-guided object detection on VisDrone using Grad-CAM++ and YOLO, with metrics and visual outputs.