Dataset from the Mapping Challenge of crowdAI: https://www.crowdai.org/challenges/mapping-challenge/dataset_files
Register and manually download train.tar.gz, val.tar.gz.
The folder structure should look like this:
mapping_challenge_dataset
|-- raw
| |-- train
| | |-- images
| | | |-- 000000000000.jpg
| | | `-- ...
| | |-- annotation.json
| | `-- annotation-small.json
| `-- val
| |-- images
| | |-- 000000000000.jpg
| | `-- ...
| |-- annotation.json
| `-- annotation-small.json
|-- LICENSE.md
|-- read.py
`-- README.md (this file)