Tensorflow implementation : U-net and FCN with global convolution

In depth machine learning resources

This repo contains auto encoders and decoders using keras and tensor flow. It shows the exact encoding and decoding with the code part.

A Numpy implementation of the dilated/atrous CNNs proposed by Yu et al. as well as transposed convolutions.

Convolutions and more as einsum for PyTorch

Semantic segmentation for road detection using fully-convolutional network

MNIST Image reconstruction using Autoencoders

Microscale 3-D Capacitence Tomography with a CMOS Sensor Array (BioCAS23)

In this project, we tend to generate some high-quality paintings using the ABSTRACT-ART-GALLERY dataset according to the DCGAN concept!

Implementation of V architecture with Vission Transformer for Image Segemntion Task

Function for transpose convolution or 'deconvolution' in tensorflow

Implementation of Variational Autoencoder (VAE) for the MNIST dataset, supporting both fully connected (FNN) and convolutional (CNN) architectures. It includes training, evaluation, and empirical analysis.

A study of the use of the Tensorflow GradientTape class for differentiation and custom gradient generation along with its use to implement a Deep-Convolutional Generative Adversarial Network (GAN) to generate images of hand-written digits.

Fully Convolutional Networks (Image Segmentation) with Resnet18 as a backbone

This repository contains a collection of deep learning projects implemented from scratch and using Keras/TensorFlow. The focus is on learning and experimenting with different neural network architectures for both foundational understanding and practical application.