Deep Learning Project - Final Project Code - Team ID: 12
- Contributors: Aditya Upadhyayula, Amrit Nidhi, Samarth Marudheri, Shreyash Kumar
-
Link for the data : https://drive.google.com/open?id=1l0s85SnJJ-3LSURkgenZ2n3L5RnfJVg5
-
Accuracy and loss plots for all models are uploaded in the plots folder.
- Below is a description of the scripts submitted in the order of workflow of the project: git a
-
convert_img.py : The scripts downsized the .jpg image to a size of 64x64. After resizing, the all images are stored as Numpy arrays.
-
GAN_train.ipynb : The script adversarially trains an autoencoder to reconstruct images from real images in the celebA dataset (trains autoGAN model).
-
GenerateFakes.ipynb : The script loads in the trained autoGAN model to generate fake/reconstructed images based on input image (pair-wise generation).
-
CNN_classifier.ipynb : The script trains and predicts a simple CNN based classifier as per the workflow described in the report.
-
DCGAN_classifier.ipynb : The script trains and predicts using a DCGAN based classifier as per the approach described in the report.