Chazie9/roboflow

Accuracy Counter

Overview of Program

This is a node program that will loop through a folder of images requesting predictions from Roboflow. When a prediction is returned it is compared against the ground truths that are provided in the _annotations file

Setup 1. cd into the program folder and run npm install 2. add images you want to test to the images folder 3. add the annotations file to the root level project and save it as a .json file 4. Add your api key to the roboflow.js file, -> line 112 5. run node roboflow.js to start the program.

TODO's

1. Convert some hard coded values to be easily changeable by the user
    - confidence interval in the axios request on line 108
    - class to be tested for accuracy 
    - imageFolder name / path
    - pull in classes from annotation file
2. Save the output of the program as a file that can be easily digested, showing the image and its accuracy so researchers can easily see what images were not accurate.