DEO Provided Human Gesture Annotations to be Used for AI Based Machine Learning Devices
Client Profile – Stockholm Based Software Firm
Located in Stockholm, Sweden, and Palo Alto, USA, the client is a leading Computer Vision based software company. Comprised of a team of Computer Vision and Machine Learning Scientists and core Software Development Engineers, the client focuses on human-machine interactions. For the last seven years, the client has been researching gesture technology and has developed a core technology framework to establish precise hand tracking and gesture recognition in 3D environments using a 2D camera that's available on any smart device.
Requirements for Human Gesture Annotations
The client needed to carefully annotate images that featured a diverse range of hand gestures, and thus required the assistance of an experienced outsourcing company. Each annotated image would be used as source input for their Machine Learning project, which would create applications that allowed for a more authentic human-machine interface.
After learning of Data Entry Outsourced (DEO) online and speaking with the sales team, the client decided to avail DEO's services due to the following reasons:
- Ability to provide meticulously categorized annotations and tagging based on behavioral classifications of human hand gestures for completely authentic descriptions.
- Ability to complete the project within the client's timeframe and budget, and provide ongoing project updates as requested.
- Prior experience in providing scientific annotation services.
Challenges Faced in This Gesture Recognition Project
The DEO team met with some key challenges over the course of this high-level hand gesture annotation project:
- DEO had to carefully differentiate between a wide range of gestures with very subtle differences and characteristics, which was a very time-consuming process.
- The client provided a large volume of images all at once, which had to be processed and delivered within a short deadline.
- The image format had to be altered before uploading them into the client's preferred file sharing program to ensure that there were no quality related issues.
DEO's Customized Gesture Recognition and Annotation Process
To successfully overcome the challenges and produce accurate annotations, the DEO team developed a customized approach to the project:
- DEO and the client engaged in a consultation to discuss the assignment parameters and ensure both parties were on the same page regarding all requirements.
- DEO assembled a team of skilled, dedicated resources with prior experience creating unique image annotations.
- The team was then trained in distinguishing the nuances of human hand gesture down to the most minute detail of movement so that the images could be annotated with complete accuracy.
- DEO's in-house annotation tool was utilized, which enabled the team to further customize the classification of the gestures with a detailed organizational framework.
- The in-house annotation tool was further customized by DEO to facilitate client's requirement.
- DEO conducted a quality control check for all images, cross-checking them against academic research of human hand gestures to surpass the expectations of the client and produce the most accurate annotations possible.
- All images were completed on time and delivered to the client via their company's file-sharing program.
End Benefits for the Client
DEO finished this complex human hand gesture annotation project successfully, and was able to deliver the following results:
- The client received all images within their requested deadline and was impressed with the involvement and approach DEO utilized with regards to the annotation categorization and descriptions.
- The client was able to use the annotated images as intended for their machine learning processes, and further develop the human-machine software interface.
- DEO was able to work on a unique annotation assignment and add another success case to the growing roster of professional achievements.