Learning-Deep-Learning

DOTA: A Large-scale Dataset for Object Detection in Aerial Images

September 2019

tl;dr: Largest dataset for object detection in aerial images.

Overall impression

This dataset addresses a specific field called Earth Vision/Earth Observation/Remote Sensing.

In aerial images, there is rarely occlusion so every object can be annotated (vs. crowd class in COCO dataset).

Other than aerial image, text region detection also involves oriented bbox detection.

Key ideas

Technical details

Notes