Vision Transfomer for classifying images

This project focuses on evaluating Convolutional Neural Networks (CNN) and Vision Transformers (ViT) for image classification tasks, specifically distinguishing between Asian elephants and African elephants.

Implements different 2D object detection algorithms. Compares accuracies.

Classifying musical pieces into appropriate genres using CNNs and Vision Transformers

An image restoration framework (Image Deraining code has been implemented) based on the Restormer model as a back-bone. This is an early idea in my "Attending to the past" research project. This model with roughly the same amount of learnable parameters shows better performance under the same training methods

Testing the Moondream tiny vision model

Benchmark for formally verifying ViTs

Deep learning pet breed recognition app

Image-based plant identification at global scale.

Transformers in Computer Vision.

Implementations of transformers based models for different vision tasks

Multitask-Learning (hard-parameter sharing) with Vision Transformers on Cifar10 & Cifar100

This repository accompanies the article entitled "Automated Classification of Oral Cancer Lesions: Vision Transformer vs Radiomics."

Based on Dosovitskiy et al., 2020. Final project for DS4440: Practical Neural Networks, Fall 2020.

Implementation of ViT with PyTorch

Final project of the Multidisciplinary course offered at Politecnico di Milano A.Y. 2022/2023

Official code for the paper "Adversarial Magnification to Deceive Deepfake Detection through Super Resolution"

Streamlit app that performs binary and multiclass classification of gravitational lensing images along with dark matter halo mass prediction.