Skip to content

sameeruddin/ACTD_scripts

Repository files navigation

Process camera trap datasets for training deep learning models

Mega Detector Annotator

1. Environment set-up

Using conda create an environment and install the libraries from the following instructions:

conda env create --file environment.yml
conda activate md-annotator
pip install megadetector

After the installation, download the mega detector model by clicking the follwoing link: MegaDetector v5b and store in the md_model directory. For detailed information, you can visit theMegaDetector User Guide.

Verify the mega detector installation by running following script:

python verify_md_install.py

Cat selfie

This script performs inference on the above Cat_Selfie.JPG image, to detect animal/person/vehicle class and predict the confidence score. The result in the command prompt should look like below:

Verify Mega Detector installation

This verification can also be performed with the Jupyter notebook (1_Generate_MD_Crop_Test.ipynb) and this notebook will generate a cropped part of the animal detected in the image.

If the execution of the above step does not print a SUCCESS message then, please check your installation and fix this issue before proceeding to the next steps.

2. Dataset generation

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors