Browse categories
Explore
Fiverr Pro
English
$
USD
I will build real time computer vision object detection for cctv security alerts
Computer Vision Engineer Real time AI detection for CCTV and cameras
Computer vision engineer with 3+ years building real-time AI detection for security and safety cameras. I have built and demoed face recognition with watchlist alerts, weapon detection, fire & smoke d...
About this Gig
Your cameras can spot a weapon, a fire, a wanted face or a fallen person and alert you the moment it happens. I build that.
3+ years building real-time computer vision for security and safety, delivered as running systems, not notebooks.
Built and demoed on real footage:
- Face recognition + watchlist: alert on wanted or unknown faces
- Weapon detection: knives, pistols, rifles
- Fire & smoke detection: forest, factory and warehouse cameras
- Fall detection: alerts only if the person stays down 30-45 s (fewer false alarms)
What you get:
- YOLO model trained on your objects and your footage
- Live pipeline for RTSP/IP, DVR/NVR or webcam, 1 to many cameras
- Alerts: screen, sound, Telegram, email or webhook + event log
- Python source code, setup guide and accuracy report
- Runs on your PC, GPU server, cloud or Jetson
How we work:
- Send a sample clip. I run a free 60-second test so you see results before ordering
- We agree on classes, cameras and alert rules
- Staged delivery: trained model, live demo, deployment
Also: people counting, intrusion zones, PPE, license plates.
Message me with your camera setup and what to detect. Fast replies, live demo on request.
APIs:
Microsoft Computer Vision AI
•
Google Cloud Vision API
Programming language:
Python
Tools:
Jupyter Notebook
•
OpenCV
•
TensorFlow
•
Colab
•
PyTorch
Frameworks:
Scikit-learn
•
Google ML Kit
•
Keras
•
PyTorch
•
Other
My Portfolio
FAQ
Will it work with my existing CCTV or IP cameras?
Yes. I connect to RTSP/IP cameras, DVR/NVR systems, USB webcams and recorded video files. Send me your camera brand or stream URL format and I will confirm compatibility before you order.
Do I need a GPU?
Not for 1–2 cameras – lightweight YOLO models run on a normal CPU. For many cameras or high FPS, an NVIDIA GPU, a cloud GPU or a Jetson device is recommended. I will size it for your setup.
Can you train the model on my own objects or images?
Yes. Send your images/videos or describe the objects. I can also collect and annotate a dataset for you. You receive the trained YOLO weights, training code and an accuracy report.
Do I get the source code?
Yes. Every package includes clean, commented Python source code, a requirements file and a setup guide so your team can run and modify it.
How are alerts delivered?
On-screen warning and sound by default. I can add email, Telegram, WhatsApp API, SMS, webhook or a dashboard. Every event is logged with a timestamp and screenshot.
What accuracy can I expect?
It depends on camera quality, lighting and data. Models trained on your footage typically reach 85–95% precision. I test on your real videos and tune thresholds to cut false alarms before delivery.
Can you deploy it on my server or edge device?
Yes – Windows/Linux PC, GPU server, cloud (AWS/GCP/Azure) or edge devices like NVIDIA Jetson and Raspberry Pi. Docker packaging is available as an extra.
Is my footage kept private? Do you sign NDAs?
Yes. Your videos and data are used only for your project and deleted after delivery on request. I am happy to sign an NDA.
