AI Camera Restaurant Monitoring
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted1 hour ago
I want to turn the video feed from my dining-room cameras into a smart, real-time assistant that keeps an eye on how well my team is serving guests. The system’s single purpose is to evaluate staff activity and overall service quality, not customer behaviour or sales.
What the vision models must recognise
• Customer wait times and overall service speed
• Staff presence at tables, pickup counters, and POS stations
• Safety and hygiene practices (e.g., gloves, hair-nets, spills on the floor)
When any of those metrics slip, the software should push the event straight onto a live dashboard and also add it to a daily summary report that I can review at closing. No audible alarms or SMS alerts are needed right now.
Scope of work
I need you to design and build the full machine-learning pipeline: video ingestion, detection/segmentation (YOLOv8, Detectron2, or a framework you recommend), activity analysis, and the business-logic layer that logs each incident. I am open to Python with OpenCV, TensorFlow, or PyTorch; the final code must run on an off-the-shelf GPU PC, with the option to migrate later to an edge device such as Jetson Nano.
Deliverables
1. Well-documented source code and pretrained weights
2. A web-based dashboard (Streamlit, Flask, or similar) that refreshes in real time and supports basic filtering by date, table, or staff zone
3. An automatically generated PDF or CSV report summarising daily issues
4. Installation guide plus a short video walkthrough that proves the system works on sample footage I supply
Acceptance criteria: at least 90 % precision on wait-time and staff-presence events in a 30-minute validation clip; hygiene infractions flagged with no more than one false positive per hour.
If this is within your wheelhouse, please outline your proposed model architecture, any pre-existing datasets you would leverage, and the estimated timeline to hit the accuracy targets.
What the vision models must recognise
• Customer wait times and overall service speed
• Staff presence at tables, pickup counters, and POS stations
• Safety and hygiene practices (e.g., gloves, hair-nets, spills on the floor)
When any of those metrics slip, the software should push the event straight onto a live dashboard and also add it to a daily summary report that I can review at closing. No audible alarms or SMS alerts are needed right now.
Scope of work
I need you to design and build the full machine-learning pipeline: video ingestion, detection/segmentation (YOLOv8, Detectron2, or a framework you recommend), activity analysis, and the business-logic layer that logs each incident. I am open to Python with OpenCV, TensorFlow, or PyTorch; the final code must run on an off-the-shelf GPU PC, with the option to migrate later to an edge device such as Jetson Nano.
Deliverables
1. Well-documented source code and pretrained weights
2. A web-based dashboard (Streamlit, Flask, or similar) that refreshes in real time and supports basic filtering by date, table, or staff zone
3. An automatically generated PDF or CSV report summarising daily issues
4. Installation guide plus a short video walkthrough that proves the system works on sample footage I supply
Acceptance criteria: at least 90 % precision on wait-time and staff-presence events in a 30-minute validation clip; hygiene infractions flagged with no more than one false positive per hour.
If this is within your wheelhouse, please outline your proposed model architecture, any pre-existing datasets you would leverage, and the estimated timeline to hit the accuracy targets.
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.