cda

A deep learning crater detection algorithm

Installation

Run "downloadimages.sh" to fetch the THEMIS mosaics from USGS via wget. Run "cropimg.sh" to split the THEMIS mosaics into tiles (requires ImageMagick to be installed)

Usage

Run generate_cratersets.py in the preprocessing folder as follows:

  $python generate_cratersets.py cfg/truth.cfg

Then train the model by running train.py in the tensorbox folder as follows:

  $python train.py --hypes hypes/overfeat_resnet_rezoom_cda.json --gpu 0 --logdir output

(note - need to change the path names in both the train.py command and also the JSON file in future commit)

Todos

Double check installation scripts run without issue Make preprocessing / training workflow simpler