Is Rojit Pokharel a web and web app developer in Kathmandu, Nepal?

Yes. Rojit Pokharel is a Full-Stack Web Developer and System Architect based in Kathmandu, Nepal, specializing in production web applications, e-commerce platforms, SaaS products, and API development using Laravel, Django, React, and Next.js.

What kind of websites and web apps does Rojit build?

Custom web applications, e-commerce and multi-vendor marketplaces, restaurant management SaaS, insurance claim platforms, travel booking systems, API services, and server monitoring tools.

Which technologies does Rojit Pokharel specialize in?

Laravel and PHP for the backend, React and Next.js for the frontend, Django for Python projects, and MySQL or PostgreSQL for databases. He also handles server setup with Nginx, Docker, and Linux.

How do I hire a web developer in Kathmandu?

Reach out through the contact page or email info@rojitpokharel.com.np. Rojit is open to freelance projects and full-time opportunities and typically responds within 24 hours.

Does Rojit work with clients outside Nepal?

Yes. While based in Kathmandu, Nepal, he works with local and international clients, communicating in English and delivering production-grade applications.

FPS: 30+· 2025

AI Referee: Football Analytics

Computer vision pipeline using YOLOv5 + SORT tracking for player detection, SSIM template matching for team classification, and pitch mapping for real-time football analytics with a Flask dashboard.

Technical case study by Rojit Pokharel — Full-Stack Web Developer & System Architect, Kathmandu, Nepal

PythonYOLOv5OpenCVSORTFlaskPyTorch

Stack diagram

Application

PythonFlask

AI/ML

YOLOv5OpenCVSORTPyTorch

Client / Problem

›Football coaches and analysts need objective, real-time player data — positions, speed, and team identity — without manual video review.
›Detecting and classifying players by team from broadcast footage is noisy: occlusion, jersey similarity, and fast motion break naive detectors.

My Role

›Computer vision engineer — built the full YOLOv5 detection and SORT tracking pipeline with OpenCV frame processing and a Flask dashboard.
›Designed the team-classification and pitch-mapping logic.

Architecture

›YOLOv5 for player detection, SORT (Kalman filter + IoU) for multi-object tracking across frames.
›SSIM template matching for team classification against known jersey templates.
›Perspective transform for pitch mapping, with a Flask dashboard serving live analytics.

API Architecture

›Flask backend streaming processed video and aggregated tracking stats to the dashboard.

Real-time Systems

›Real-time video inference pipeline sustaining 30+ FPS with detection, tracking, and classification per frame.

Deployment

›GPU-backed inference environment running the PyTorch pipeline, with the Flask dashboard served over HTTP.

Performance Optimization

›Frame-by-frame YOLOv5 inference optimized to sustain real-time rates.
›SORT tracking keeps player identities stable between detections, reducing per-frame compute.

Problems Encountered

›Occlusion and similar jersey colors confuse both detection and classification.
›Tracking IDs must remain stable when players cross paths or leave frame briefly.

How I Solved Them

›Combined detection with SORT tracking so identity persists through brief occlusions.
›SSIM template matching on player crops assigns team affiliation before downstream analytics.

Results

›End-to-end AI football analytics pipeline: detect, track, classify, and map players in real time.
›Demonstrates production-grade computer vision with YOLOv5, OpenCV, SORT, and PyTorch.

Lessons Learned

›Tracking is the bridge that turns detection frames into useful sports analytics.
›Template-based classification is robust when lighting and jerseys are controlled.

Building an AI Referee: Computer Vision for Football Player Detection and Tracking

Build an AI-powered football analytics pipeline using YOLOv5, SORT tracking, SSIM-based team classification, pitch mapping, and a Flask dashboard.

Answers

Read the Answers

Direct answers to questions about Rojit Pokharel and the technologies used across these projects.

Want a production system like this?

I build production web applications end-to-end. Let's discuss your project.

Get in Touch