adityakulkarni3007's repositories
.bash_settings
A set of useful custom bash settings
Latent_Interpolation_DCGAN
The project aims at traversing through the latent space of the Generator to produce images that morph into eachother. The network used for this project is a Deep Convolutional GAN. The celebrity faces data-set is used as training data and the fake images produced are visualized.
Controllable_GAN
The goal is to explicitly control the outputs of a GAN. A Deep Convolution GAN and a Classifier are trained from Scratch on the CelebA Dataset. The Classifier and Generator are frozen and the gradients are used to update the latent space to generate images from the given class
UNET_ImageSegmentation
The project aims at producing segmented images of cars. The model used is a slightly modified version of the proposed Convolutional Neural Network in the U-Net paper. The model was able to achieve a pixel accuracy of over 99.4% and a dice score of over 98.7%.
Mini_Pojects
My first mini projects
Facial-Age-Predictor
A Convolutional Neural Network with the goal of predicting the age of a person(as an integer). The model is a pre-trained VGG16 model which is fine-tuned for the faces dataset, which contains a limited number of input images from 100 different ages.