The Swiss company offers a state-of-the-art imaging solution that provides instant analysis of food waste. With artificial intelligence as the foundation, client leveraged food image data to raise awareness and tackle food waste for major restaurants and hotels worldwide.
Food waste assessment solution provider had footage of garbage throwing activity by way of camera captured images. The client was looking to identify, categorize and label thousands of food product images as a preparatory step towards its data annotation activities.
Hitech was approached to study and label customer waste and kitchen waste food images according to pre-defined food categories. The labeled images were further to be used to train client’s machine learning models to see and interpret visual information just like humans.
- Hiring and training data annotation specialists experienced at working on food and beverages information
- Product images were not recognizable when there were multiple food items in the food tray
- Understanding, working and labeling a huge range of unfamiliar European food products
Identified, categorized and labelled thousands of food images from across hotels and restaurants to prepare extensive training data to be fed into machine learning models.
After initial assessment of food product images on client’s software, team of data annotators at Hitech documented a workflow to expedite image labeling. The images were annotated through a five-step process:
- Labeling and Segmentation
- Data annotation specialists labeled food product images against an existing master repository of name of food items in form of a drop-down menu in client’s portal. The task involved going back and forth with the visual data of the garbage throwing process and reviewing the items from various angles for accurate identification of food objects.
- Tagging and segmentation of images against client defined product list and standards of restaurants and hotels.
- Fuzzy or low confidence images were routed separately to be validated by the client.
- Audit and review
- As part of the quality check process, the client conducted a parallel image labeling activity in-house for the same set of images. Resultant data was used to validate anomalies in the labeled images submitted by Hitech annotators.
- The data deviations were used by the client for QC and training purposes.
- Erroneously labeled or missed out images were re-labeled appropriately to ensure accuracy.
- Automated upload of labeled food images on client portal to serve as machine training data.
- Reports/dashboards with insights on number and type of images annotated.
- Additional analytical insights by factoring in numerous metrics such as the day, time and type of food wasted, total food wasted against total quantity prepared etc.
Secure Login to Client portal through web-browser
Accurately annotated food waste across categories fueled machine learning models.
Thousands of images annotated at speed, fast tracked food waste analysis.
Factoring in multiple metrics helped derive actionable analytical insights.