I will build computer vision and yolo object detection
AI Engineer and Product Architect
Level 2
Has met high performance criteria and has a proven track record for meeting client expectations.
About this Gig
I build custom computer vision systems that detect, read and classify what matters in your images, video and live camera feeds.
Proven in the real world:
- Parking access system: YOLOv8 vehicle detection at 99.2% on a live RTSP camera, number plate OCR at 96% mAP50, registry lookup in under 2 seconds, gate control and a Flask dashboard. Deployed at a live facility.
- RoadEye: YOLOv5 on 4K drone video to detect and classify road defects, with GPS location for each one.
- DermaSense: a desktop app that detects skin lesions with YOLOv5 and adds a GPT-3.5 consultation layer.
- Invoice OCR: a Python and PyQt app that reads invoices with Tesseract or EasyOCR and exports CSV or JSON.
What I can build for you:
- Custom YOLO object detection trained on your images
- Number plate, text and document OCR
- Real-time video and camera pipelines
- Dashboards and APIs around the model, with Flask and Next.js
Please send: sample images or video, the objects you want detected, where the model will run and any speed or accuracy targets. I will tell you honestly what your data allows before we start.
Expertise:
Image processing
•
Classification
•
Object detection
Programming language:
Python
My Portfolio
FAQ
Can you train YOLO on my own dataset?
Yes. I train or fine-tune a YOLO model on your labelled images and the classes you need, then evaluate it and integrate it into your app or API.
Can you process live camera or video streams?
Yes. My parking system runs on a live RTSP camera. Speed depends on your hardware and accuracy targets, and I confirm that before starting.
Can you read text or license plates?
Yes. I have built plate OCR for Arabic and Roman numerals at 96% mAP50, and I can add OCR to other detection use cases.
Can you label my images?
I can advise on how to label them. Large labelling jobs are best agreed as a custom order first.

