A Streamlit aaplication to distinguish between AI-generated and real images. Utilizing the TensorFlow framework, the core model is based on EfficientNetB0, which has been fine-tuned on a custom dataset containing both fake and real images.

Fairness Auditing in Dermoscopic AI: Quantified a 55% FNR disparity based on Anatomical Localization (Spurious Bias Audit). Focus on Disentangled Representation learning for Equitable AI.

Mask Monitoring System

Brain tumor MRI classification using deep learning models such as ResNet18 and EfficientNet-B0 to identify the most effective architecture for accurate tumor detection.

DeepFake Detector using a dual-branch neural network combining EfficientNet-B0 and handcrafted features. Trained in two phases with early stopping and deployed via Streamlit for real-time inference. Supports demo images and custom uploads with confidence scoring.

Semantic segmentation of road scenes using U-Net with EfficientNetB0 on the CamVid dataset.

This project focuses on pneumonia detection using chest X-ray images and deep learning. It leverages transfer learning with the pre-trained VGG16 model to classify images as Pneumonia or Normal, aiming to assist in fast and accurate medical diagnosis.

Efficient way to detect if a face has a mask on. Used EfficientNet-B0 model architecture.

INR Denomination Recognition is an image classification project

Implementation of a neural network for solving the problem of mushroom classification based on the TensorFlow library and the pre-trained efficientnet/b0 model.

Real-time Finnish traffic sign detection and classification using synthetic data, transfer learning, and advanced deep learning techniques.

This application predicts the name of a country (or countries) based on an input flag image. It uses advanced image processing techniques and deep learning models built with PyTorch to classify flags accurately.

Ship classification for the minor project

A project on building deep learning classifier to classify playing cards

A deep learning project for classifying 130+ fruits using EfficientNet, ResNet, and MobileNet with custom augmentations and SE blocks. Built on the Fruits-360 dataset.

EfficientNetB0-based skin lesion classification on the HAM10000 dataset using strong data augmentation and class weighting to address class imbalance. Includes training pipeline, evaluation metrics, and comparative analysis with previous CNN architectures.

Deep learning-based system for detecting synthetic facial manipulation in low-illuminance environments

This repository contains the implementation and documentation of deep learning models designed to classify the stages of Alzheimer's disease using MRI images. The project explores the effectiveness of pre-trained convolutional neural networks (CNNs) such as VGG16, ResNet50, and EfficientNetB0 in detecting and categorizing neurodegenerative disease

🧠 Detect brain tumors in MRI images using an efficient deep learning model with transfer learning and explainable confidence analysis.