- Case study ↓
- AI/ML
- Computer Vision
Kitchen Monitoring That Maps Where the Work Happens, Automatically
Computer vision that turns overhead kitchen video into heatmaps of high-activity zones, so you can focus hygiene checks where they matter and plan a better layout.
The demo
See UV Scan / Hygiene at work
- Industry
- Computer Vision
- Input
- Overhead kitchen camera video
- Output
- Activity heatmaps, Hygiene risk zones
- Built with
- Python · OpenCV · Gaussian Mixture Model · Heatmaps
The problem
You can't clean what you can't see
Monitoring how people use a kitchen by hand is slow, error-prone and can't adapt as the day changes. Nobody knows which areas see the most activity in a shift, so hygiene effort is spread evenly instead of where it's needed.
Kitchens are also hard for cameras: changing light, shadows and reflections from steel surfaces make movement difficult to pick out with traditional techniques.
The solution
What UV Scan / Hygiene does
Activity heatmapsSee which zones are used most, generated from the video
Shadow and noise suppressionReflections and lighting changes don't show up as activity
Sensor-freeWorks from a camera alone and scales to more kitchens
Hygiene audit supportFocus cleaning checks on the busiest, highest-risk areas
Layout and ergonomicsSpot crowded zones and plan stations around real use
Technology
How it’s built
Every step, from the overhead kitchen camera video to the results.
Step 1
Input
- Overhead camera video
Step 2
Background subtraction
- Gaussian Mixture Model
- OpenCV
Step 3
Clean-up
- Thresholding
- Dilation
- Morphological processing
Step 4
Output
- Activity heatmaps
- Hygiene risk zones
Who it’s for
- Restaurant and QSR chainsWho want consistent hygiene across every location.
- Food production sitesShowing auditors where activity and hygiene risk are highest.
- Kitchen designersPlanning layouts and stations around how space is really used.
Case study
UV Scan / Hygiene: the full case study
- Overview
This project uses computer vision and machine learning to analyse human activity in kitchens from fixed overhead camera footage. It extracts patterns from daily kitchen work to help optimise hygiene, layout design and use of space.
- Problem
Monitoring human interaction in kitchens by hand is inefficient, error-prone and can't adapt in real time. Lighting, shadows and reflections make it hard to capture reliable movement patterns with traditional techniques.
- Solution
ML Bench built a solution based on a Gaussian Mixture Model (GMM) that performs dynamic background subtraction on overhead video. The system learns the scene and segments human movement from it.
The segmented movement is turned into heatmaps that show high-activity zones and potential hygiene risk areas, supporting hygiene audits, layout optimisation and ergonomic design. It is built in Python with OpenCV and NumPy.
- Challenges & solutions
Shadows and reflections were detected as movement.Retrained the model with extra data focused on reflection-heavy sequences.
Noise in the detected foreground made heatmaps patchy.Used optimised thresholding and dilation to restore coherent heatmaps.
Tuning for clarity slightly eroded the edges of people and objects.Advanced segmentation is planned as future work to recover edge detail.
- Results
- Heatmaps that clearly highlight the kitchen's most-used zones
- Visual insight into human activity without sensors or wearables
- A foundation for future improvements in detection accuracy and space optimisation
Have Questions?
Find the right solution
for now
No. It separates movement from the background to build heatmaps; it doesn't recognise who anyone is.
Why ML Bench
Why teams choose ML Bench
Built for real-world footageWe build computer vision for real cameras and scans: retail, sports, healthcare, logistics and safety.
Starts with the data you haveExisting cameras, recordings or image archives are usually enough for a first proof of concept.
Custom to your setupWe adapt the model to your camera angles, image quality, classes and workflow, not the other way round.
One team, pilot to productionData, model training, dashboard, API and deployment handled by the same team.
Next AI/ML project →
Cell Segmentation
AI that segments, classifies and measures every cell in microscope images, with per-cell and per-image metrics.
Free assessment
See Your Kitchen as a Heatmap
Send us a short video from your kitchen and we'll show you what the AI picks up, free.
- UI/UX Design
- Web Development
- Mobile Apps
- SEO
- AI/ML Solutions
