Audience analytics$975 USD
Audience Measurement
Measure viewing time and gaze-based attention around a display, with facial expression estimates. Build the results into your own analytics.
Independent computer vision engineering
Computer vision.
Working engines.
Full source code.
I’m Antal. I build C++ and OpenCV engines you can try in your browser, understand through the documentation, and integrate into your own product.
Commercial use · Digital delivery · Built for independent integration
An introduction to my work and the engines you can build on. Watch on YouTube ↗
Inside the featured enginesC++OpenCVNeural networksWebAssembly
Engineering since 2012Industry and personal projects
Try before you buyLive demos, no account needed
Read the implementationPublic engine documentation
Build on your termsSource you can study and adapt
The project collection
A working start for your next feature.
Choose the capability or dataset you need. Evaluate the demo or sample, review the contents, then build on the work.
Identity & verification$325 USD
Liveness Detection
Add action-based liveness checks to your face verification flow. Ask for blinks or head turns and check the response.
Identity & verification$325 USD
Face Recognition
Detect, track and re-identify faces in video. Start with a working C++ engine and adapt the matching to your application.
Identity & verification$325 USD
Face Anti-Spoofing
Evaluate a neural-network approach to detecting photo and screen presentation attacks in a face verification workflow.
Gaze & attention$325 USD
Gaze Tracker
Estimate head pose and eye gaze direction from a webcam, with facial landmarks and eye openness measurements.
Expression analysis$325 USD
Facial Expression Recognition
Classify visible facial expressions in a camera feed and explore the estimates in your own visualization or analytics.
Creative vision$325 USD
Virtual Makeup
Map your own 2D makeup designs onto a tracked 3D face mesh, with a working browser demonstration.
Research prototype$325 USD
Webcam Heart Rate Monitor
Explore contactless pulse estimation from subtle facial color changes using remote photoplethysmography (rPPG).
Research prototype$99 USD
Synthetic Webcam Scene Dataset
17,280 prompt-generated webcam images across 12 categories, with original sheets, prompts and companion source code. For research and training experiments; no trained model included.
The engine packages above include live demos, public code documentation and downloadable source. The dataset has a free image sample and a companion case study. Self-serve purchase; support and custom integration are not included. What to know before buying.
From demonstration to implementation
The code is part
of the product.
A working demonstration is the starting point. The featured engine packages give your team the implementation and materials to understand it, compile it and make it your own.
- Inspect the logic. Follow the processing pipeline in C++.
- Adapt the behavior. Change how the engine fits your application.
- Control your deployment. Build with the dependencies and environment your project needs.
- Ship your own product. Commercial integration is permitted under the license.
Inside a featured engine package
01 C++ source codeThe engine implementation you can inspect and modify.
02 Models & assetsPretrained neural networks and the supporting resources.
03 Documentation & build guidesCode reference and Emscripten compilation instructions.
04 OpenCV WebAssembly moduleA precompiled dependency for the browser build.
A self-serve path
Evaluate. Download. Build.
Try your actual use case.
Open a live demo. Check the result with your camera, lighting and distance. Read the engine’s capabilities and limitations.
Know what you’re getting.
Review the public documentation, package contents, license and support policy. Confirm that your team can handle the integration.
Make it part of your product.
Purchase through the shop and receive your download links. Compile, adapt and validate the source in your own environment.
A good fit for technical founders, developers and engineering teams. You should be comfortable compiling C++ and working with OpenCV. The projects are engineering foundations; production integration, testing and ongoing maintenance are yours to manage.
Get to know the work
Start with a free project.
Explore a smaller project before choosing an engine. The same license applies to free downloads.
Free download
Person Pixelizer
Neural network–powered tool that segments and pixelates people in real time to protect privacy
Free download
Color Checker
WebAssembly app clustering camera pixels to reveal the scene’s dominant colours and HEX codes
Free download
Facemask Detector
This model detects the wearing / non-wearing or improper wearing of a medical mask
Free download
Pothole Detector
This model detects potholes on the road
Free download
Handgun Detector
This model detects handguns
Free download
Anti-Spoofing · YOLOv4
An older standalone model for face presentation-attack experiments, with a Windows demo to evaluate it.
Free download
Pupil, Iris & Sclera · YOLOv4
Detect pupil, iris and sclera bounding boxes in cropped eye images. Includes a link to the Windows demo.
The engineer behind the engines
Built by me.
Ready for you to build on.
I’m Antal Zsiros, a computer vision engineer in Hungary. My work spans image processing, embedded vision, automotive navigation and face analysis, with projects going back to 2012.
Antal.AI is my independent engineering workshop. I develop these projects alongside my professional work because I enjoy solving the problems behind them. Clear documentation, working demos and honest boundaries are how I make that work useful to you.
Antal Zsiros
Clear expectations
Before you make
it part of your product.
Source access, commercial use and integration responsibilities, explained up front.
What exactly am I buying?
A self-serve digital project for developers. The engine packages include C++ source code, pretrained models, documentation and build instructions. These are foundations for your own application; you handle integration and production validation. Standalone model downloads have their own package contents.
Can I use the code in a commercial product?
Yes. You can modify the software and incorporate it into your own product that adds substantial value. You may sell that product, but may not publish the supplied source code or resell the software as a standalone library, SDK, toolkit or template. Read the full software license (PDF).
Will it work with my platform?
The featured engines use C++ and OpenCV; their browser demos are compiled with Emscripten to WebAssembly. Native desktop, server, mobile or edge deployment requires the appropriate OpenCV build and project configuration. A browser demo does not establish compatibility or speed on your target device.
Is support or custom integration included?
No. These are self-serve products for people who can work with C++ and OpenCV. Installation, integration, optimization and custom development are not included. Updates may be released at my discretion, with no maintenance schedule or response-time guarantee. Read the support policy.
Earlier engineering work
Research & prototype portfolio.
Explore the research & prototype portfolio
These project pages document earlier work. They are not offered as complete, supported systems in the shop.
Driver Monitoring Driver Fatigue Detector Driver Distraction Detector Face Occlusion Detector Road Traffic Analyzer Face Tracker Eye Landmark Detector Missing Object Detector Head Gaze Tracker Eye Gaze Tracker Online Proctoring Extended Face Detector Seatbelt Detector Insect Detector Mobilephone Detector Fire Detector
Find your starting point.
See what works for your use case before you commit.