Case Study on Bounding Box & Annotation Services

The Customer

The client is one of the leading providers of visual search and image recognition technology. It offers visual tech APIs to businesses with large image databases who want to simplify search inside their own website or mobile platforms.

The client was looking for a partner who could provide annotation and bounding services for 1 million images of apparels and accessories.

The client needed a team to assist them with recognizing apparels and accessories from within their Customer Relationship Management (CRM) software, highlighting each product by drawing a box around it, and tagging each product using a pre-defined list provided by the client.

The Challenge

The client had insufficient manpower to complete the task on their own, and was also looking for an affordable solution to match their budget. There were also several challenges that would require specific expertise and organization, including:

• Inefficiently Categorized Products: The product database had incorrect category mapping, potentially leading to complications with tagging and recognition
• Required Human Intelligence: Automated product recognition software was not be sufficient for this task, which required visual recognition of products, associating products with one or both genders, and other related distinguishing characteristics which are not possible with any software
• Short Deadline: The project needed to be completed in 30 days, which was a short timeframe given the human intelligence and categorization issues, as well as the large volume of images

The Solution

To respond to the challenge, 3 resources were assigned exclusively to this project, and trained in the customer's CRM software. The team sent a few samples to the client so that they could gauge our team's performance. Once the samples were approved by the client, the team started working on the project as per the client's specifications. Products were then categorized into "buckets", to ensure efficiency and accuracy during the tagging process.

New categories were added for products that did not match any existing categories using the pre-defined tags provided by the client. Products that were difficult to categorize and tag were given extra attention to ensure correct tagging. Flatworld Solutions communicated every 2 days with the client to and update them on progress.

The Results

Despite the difficulties, the project was completed within the 30 day turnaround time. Client reported an accuracy rate of 100%, with no errors found with tagging or categorization and Flatworld Solutions was also able to successfully deliver within the client's budget.

Since the first project was successfully completed, the client has come back to Flatworld Solutions on several occasions for further projects. Although the client originally contacted Flatworld Solutions for help with a single project, they have since continued to use Flatworld Solutions for their ongoing image tagging and data management efforts.

If you have a large database of images and need help with tagging, organization, management, or any related issues, contact Flatworld Solutions today for affordable and efficient image management services.