cs4100-project

CS 4100 Final Project!

Local Setup: Get Started:

This project uses python 3.12 because it is the latest version supported by tensorflow, which we need because our data format is in TFRecord format.

Necessary installs

• Install the requirements.txt

Populate data locally

• Download the objects.tfrecord file from https://drive.google.com/file/d/1YKlQE8oCPcr7lGOFeSl_UrMgt6HroPiq/view and put it in the ./data folder
• Run python processing/process_data.py to fetch the data for the data
• This will populate the data/img folder with the training images and data/chords/labels.json with a map from image file name to chord number

Running the CNN

• Run python models/cnn.py
• The terminal will output the epoch results and some other logs
• The result of model will be saved to a folder with the name of format model_ACCURACY_EPOCHS
  • In this folder will be the model and the data vis for the model and training results

Running the Landmarking NN

• First populate the data
• Run python models/nn.py to run the neural network
• The accuracy of the network will be printed in the console