coputer-vision

🦺 Real-time AI safety monitoring for construction sites detects people, tracks movement & analyzes crowd trajectories using YOLOv8 + OpenCV

This is a flood detection image classification project. The goal is to compare multiple deep learning architectures for the same task and provide a clear, interactive way to test them.

An end-to-end computer vision pipeline for football analytics. Uses YOLO, ByteTrack, and OpenCV to automatically detect players, track ball possession, compensate for camera motion, and calculate real-world speed and distance from raw match footage.

AI-powered touchless computer control system using OpenCV, MediaPipe and SVM for real-time hand gesture and head movement interaction.

YOLOv8-based car detection and monocular distance estimation evaluated on the KITTI dataset

Plant disease classifier built with MobileNetV2 transfer learning, fine-tuning, TensorFlow, and Streamlit.

A computer vision project that detects and tracks human body keypoints in images and video using 'Digital Image Processing techniques.

A computer vision project that performs real-world object detection and tagging using Connected Component Labeling (CCL) — applied to various real-world images including cars, coins, binary images, and more.

A pure JS code to train/run a model to recognize hand written digits, no library of either external or build-in is used.

AI-assisted MRI ChangeGuard for deterministic image-quality assessment, longitudinal study comparison, experimental CNN inference, and auditable release validation, developed with IBM Bob.

EyeNet is a camera-first, real-time surveillance system intended for campus / institutional environments. It continuously processes a live video feed and converts computer-vision detections into actionable incidents: persisted audit records, real-time dashboards, and automated notifications.

🚗 🇧🇩 Vehicle Detection & Classification in Bangladesh Highway Surveillance | YOLO11x Fine-Tuning

A multimodal deep learning framework for real-time detection of deepfakes and AI misinformation. Fuses semantic text analysis (NLP) with frequency-domain video forensics (FFT) to generate a unified Trust Score.

AI-powered halal food detector that analyzes product photos using OCR, logo detection, and databases to determine halal status with confidence scoring.

A real-time Human-Computer Interaction (HCI) system that transforms standard webcam feeds into a low-latency, touchless mouse using computer vision, MediaPipe hand landmark detection, and native OS automation.

This project uses YOLOv8 and OpenCV to perform real-time object detection using webcam input.