Skip to content
Code for Methods in Ecology and Evolution paper: "A Convolutional Neural Network for Detecting Sea Turtles in Drone Imagery"
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
DukeTurtle_info.h5
DukeTurtle_info.json
DukeTurtle_test.mat adding full stack of processed images, model definition, and model we… Dec 5, 2018
DukeTurtle_test_cnnClass.mat
LICENSE Create LICENSE Nov 18, 2018
README.md
cnn_predict_stack.py adding full stack of processed images, model definition, and model we… Dec 5, 2018
data.py adding full stack of processed images, model definition, and model we… Dec 5, 2018
run.sh

README.md

cnn_sea_turtle_detection

DOI

Code for Methods in Ecology and Evolution paper: "A Convolutional Neural Network for Detecting Sea Turtles in Drone Imagery"

Paper can be accessed at: https://doi.org/10.1111/2041-210X.13132

Using this code:

Running run.sh in bash will run the full turtle detection workflow.

Notes:

  • Python 2.7 is required and nonstandard python packages necessary are: numpy, scipy, keras, tables, and hdf5storage
  • This setup runs on preprocessed imagery contained in the .mat file. Full turtle image data along with labels for independent machine learning development can be found at doi:10.5061/dryad.5h06vv2

File Details:

  • data.py
    • defines utility functions for model creation and matlab ingestion functions
  • cnn_predict_stack.py
    • run prediction on processed images
  • DukeTurtle_info.h5
    • Trained model weights file
  • DukeTurtle_info.json
    • Model definition file
  • DukeTurtle_test.mat
    • processed and tiled RGB image data that is fed into the model. Training / validation split is 85% train / 15% validation
You can’t perform that action at this time.