A workplace hygiene solutions company approached AI Superior with a unique task: to create a system capable of autonomously identifying when an area needed cleaning, reducing the need for human intervention. This project consisted of two phases, involving the development of object detection models to identify objects and people within workplace environments.
Enhancing Workplace Hygiene with AI-Powered Object Detection
Summary
Challenge
The challenge was to develop a system that could autonomously detect and assess the cleanliness of workplace areas, recognizing objects and distinguishing clean from dirty or messy areas. Additionally, the project aimed to detect people within these areas while respecting privacy concerns, ensuring that images with people were not processed or stored.
Solution by AI Superior
In the first phase, AI Superior developed a robust object detection model capable of identifying and classifying various workplace objects (including bins, cleaning products, food waste, cups, glasses, backpacks, handbags, and more) in zenithal photos. The model also categorized areas as clean or messy based on dynamic rules.
The second phase focused on developing a model to detect people within workplace areas. Precision was crucial to avoid processing or storing images containing identifiable individuals, respecting privacy concerns. AI Superior implemented a person detection model that could identify partial representations of individuals, such as a foot or hand, without capturing recognizable facial features.
Outcome and Implications
Despite limited publicly available zenithal photo datasets, AI Superior successfully developed a highly precise model for identifying objects within workplace environments and distinguishing clean from dirty or messy areas. This system’s ability to detect and classify objects provided actionable insights, facilitating timely cleaning interventions while the privacy-focused person detection feature ensured responsible and ethical handling of sensitive information.
English 
German