Image Annotation & Tagging for the World’s Largest AI-Powered Image Recognition Engine
The Client – AI-enabled Service Provider for Retailers
Established in Paris, France in 2015, the client is constructing the world’s largest AI powered image recognition engine for digitizing supermarket shelves. They provide real-time analytics for stores using robots, smart phones, drones, and shelf cameras for on-shelf inventory management and merchandising optimization. Their solution improves the customer experience, operation costs, and brand collaborations using advanced deep learning with over 96% accuracy.
Image Annotation and Tagging Requirements to Feed AI-based Machine
The client wanted to highlight the supermarket merchandise with bounding boxes and annotation using the appropriate industry terminologies that described products for identification and information. With only a 14-day deadline, the client needed an outsourcing data entry partner who could effectively handle a massive volume annotation project without sacrificing quality.
Learning of Data Entry Outsourced (DEO) through an online inquiry, the client contracted our services due to the following:
- Ability to handle a massive scale project and input volume of 1.4 million images.
- Reputation as an industry leader in data entry accuracy and quality, with a team of certified and skilled FTE technicians.
- Proven experience working in CRM management, bounding box and annotation solutions for B2C industries.
Key Challenges Addressed by DEO’s Data Entry Team
DEO dealt with a few challenges while working on this high-volume image bounding box and annotation project:
- The tight deadline, including 1 million image annotations to be completed in 14 days with QC processes, presented extreme time management challenges.
- Onboarding a 115-member team and facilitating coordination across all resources was difficult given the lack of time that could be spent preparing.
Image Annotation and Bounding Box Process for This Project
DEO designed a step by step process to help overcome the challenges and ensure that this annotation project could be well managed:
- DEO pulled together an experienced team of 110 data entry technicians and 15 Quality Control Leads to work exclusively on this project for 14 days.
- DEO began by accessing the database of images in the client portal, which contained unique IDs for each product image.
- The product in each image was identified, and the bounding boxes were created around the images and their product content.
- Each product within the bounding boxes were appropriately annotated with the correct terminology based on similar industry-based perishable and consumable tags and keywords.
- Quality Control was conducted across image batches and upon verification of expected quality benchmarks, was submitted into the client's CRM system.
- DEO was able to complete annotation for 1.4 million images in the 14-day deadline.
Value-added Steps by Our Team
DEO’s team added complimentary additional classification and keywords to each image during the annotation process beyond the client’s initial request to simplify identification and retrieval.
Key Project Benefits from This Image Annotation Project
DEO was able to successfully manage this massive scale annotation project with the following key takeaways:
- The client was extremely relieved and pleased to have found an experienced data entry company who could provide 115 experience resources for a two-week project so they could meet their business obligations and devote internal resources to more urgent tasks.
- Being impressed with our performance and commitment, the client procured another contract for DEO’s services that would position us as the full-time data entry service provider for all of their relevant projects going into the future.
