Annotation Photos for the Recognition System

A startup approached us seeking improvements in their recognition system, particularly for intricate objects like individual strands of hair, various hairstyles, accessories, piercings, etc. Solving this problem required intricate annotation involving object matting and semantic segmentation. Additionally, assessing the presence of specific details as a percentage of the overall image area was crucial. None of the available annotation tools met these requirements as outlined in the instructions, prompting the startup team to explore alternative solutions, leading them to Training Data

Given the inadequacy of segmentation tools, we assembled a team of 25 individuals and conducted a training program on using a graphic editor tailored to the project’s needs. Following the two-day training, assessors achieved a high level of proficiency, resulting in a remarkable 95% accuracy in determining and classifying areas of interest within photographs

• The customers successfully improved the existing object capture technology within their app
• Aggregator ratings (Google Play and App Store) increased from 3.8 to 4.5
• The customer’s team has undergone training for autonomous operation with subsequent datasets