C

computer-vision

An open-source computer vision framework for wildlife image analysis, featuring state-of-the-art models for species classification and detection.

AR-примерка очков в браузере. MediaPipe FaceLandmarker (478 точек лица) + OpenAI GPT-4o Vision + DALL-E 3. Без установки, без сервера — работает локально в браузере.

🥇 MedViT-X: Vision Transformer for Medical Image Intelligence

A pipeline for curating and sanitizing large-scale image datasets.

Follow me - Vision is a vision-based control system designed to automate the navigation of remote-controlled objects like drones or cars. Using computer vision, the system detects a target object and generates commands to autonomously follow the subject in real time. Ideal for applications where a vehicle needs to track and follow a designated guide without manual intervention. For example, it could be used to help an elderly person by having a shopping cart follow them around a store.

This project is part of a larger, future initiative called Follow Me !

Proyecto desarrollado para parqueadero con el fin de detectar la cantidad de salida y entrada de vehiculos (motos y carros) mediante cámaras.

[2023] Interactive thermal DJ system translating body heat into music control. MLX90640 IR camera detects thermal events that trigger tempo changes, scratching, and track selection via automated Mixxx control with real-time thermal visualization.

A GUI-based Python utility engineered to remove image backgrounds using a suite of AI models. It streamlines testing by processing one image with multiple algorithms (rembg, mediapipe) and saving separate, transparent outputs for easy side-by-side comparison.

Real-time YOLOv7/YOLOv7‑Tiny object detection on low‑powered Jetson Orin Nano using TensorRT and a multi‑threaded ROS 1 pipeline for edge robotics, drone monitoring.

A CBIR and classification system that uses both representation learning and handcrafted features to build the image descriptors.

Population density tracker using Computer Vision.

3D Hand Pose Estimation using Multi-View CNNs