← All project workComputer vision / Pier 60

Turning a changing rental layout into data.

A computer-vision demo that detects beach rental units, classifies occupancy and maps their positions as the layout changes.

Project
Pier 60
Contribution
Founder-led work under Revenue Optimization Group, LLC, before VexWorks was formed.
Current stage
Working demo
01 / Contribution

The problem and the work.

Beach rental units move, and a fixed layout does not describe what a camera sees throughout the day. The project aimed to identify individual units, determine whether they were occupied and keep track of their positions.

Joshua Sohonyay built this work under Revenue Optimization Group, LLC, before forming VexWorks. The scope included training-data preparation, custom models, labeling tools, positioning logic and a working demo.

  • Data and labeling toolsManual and model-assisted labeling apps, plus curated training datasets.
  • Detection and occupancyCustom-trained YOLO models to identify umbrellas, cabanas and whether a rental unit is occupied.
  • Positioning and evaluationGeometry and custom logic to map units as they move, with evaluation against recorded occupancy.
02 / Implementation

Inside the system.

Prepare the camera data

Manual and model-assisted labeling tools supported dataset curation. The resulting training data represented the rental units and occupancy states the model needed to recognize.

Train models for the rental task

Custom YOLO models detected rental units in camera views and classified occupancy. The work targeted this rental setting rather than relying only on general-purpose object labels.

Map detections back to the layout

Geometry and custom logic associated detections with positions in the rental area. That mapping accounted for units being moved as the layout changed.

Compare predictions with recorded occupancy

The implementation compared predictions with recorded occupancy to evaluate detection and occupancy errors.

Pier 60 technical notes

The working demo covered detection, occupancy classification, positioning and supporting data tools. The project ended before full deployment. A demo is evidence of an implemented pipeline; it does not establish production deployment or a quantified business result.

03 / Delivery

Where the work stands.

Working demo

The demo and implementation were built. The full deployment was not completed because funding was not available to finish the project. This is earlier work by the founder, rather than a completed VexWorks deployment.

Project notes updated .

04 / System outline

From camera view to rental data.

The pipeline combines detection, occupancy classification and position mapping. Evaluation compares the output with recorded occupancy.

Implementation outline
  1. Camera viewImages of the current rental layout.
  2. Detection & occupancyCustom YOLO models identify units and their state.
  3. Position mappingGeometry and logic locate units as the layout changes.
  4. EvaluationCompare the model output with recorded occupancy.
A simplified view of the implemented system.

Bring us your version of the problem.

We can scope a project around your tools, data and the outcome you need.

Discuss a project