A dataset to detect UI elements in geoguessr videos!
This uses Anaconda to create the package, and installs requirements from requirements.txt. The environment name is defined in the ./manifest
file.
cd scripts
./create_env.sh
Copy images (~26K total) from /shared/gbiamby/geo/screenshots/screen_samples_auto/
.
Alternatively you can set img_dir
in the dataset_config section of your config file(s) to point to the above path and skip the copying step.
cd ./tools
python train_geo.py --config_file ../configs/mmdet_faster_rcnn.resnest.yaml