2D Object Detection Dataset captured by Fisheye lens camera in Dongseongro, Daegu, South Korea We have released the first set of tasks in our Google Drive (link)
This dataset consists of 5016 images and 2D bounding box annotation files. We provide 2D boxes for 9 classes: pedestrian, rider, person sitting, bicycle, motorcycle, car, van, truck, bus
An example image is in Fisheye-Dongseongro/ReadMe/, also shown below.
The data format is based on YOLOv3 labels.