Deep Learning Algorithm for Identity Management
Smart Object Solution for Benjakate. Benjakate Technologies was providing a software solution to Microsoft. Talentas Technology supported Benjakate in developing the proof of concept and executing the project in a short duration.
Business Challenge
Benjakate required new identity management solution to cope with new challenges due to the inherent nature and requirements of IoT, where the identities of large amounts of heterogeneous smart objects need to be managed. The proposed solution needed to be adapted to the envisioned scenarios allowing more flexible sharing models, as well as volatile and dynamic security associations between entities, while privacy is still preserved, keeping data minimization as a main concern.
Solution
Talentas developed an essential feature that provides support for finding the IoT devices in order to be addressable, named, and finally discovered. The most important part of machine learning is training the dataset. Talentas also developed a tool which could train the program on real world video data which discovers the objects frame by frame in each video. Based on this training data, our AI program was written in python.
Key Technologies & Platforms
- Python for Image/Video Processing
- .NET & Java Script (MVC) for training dataset
Software Methodology
- Rapid Prototyping
Outcome
The Smart Object Solution was built in 4 weeks with following features:
- Identifying the appearance of object in a video or stream of videos
- Counting objects
- Providing Real-time streaming analysis
- Providing Highly responsive alerts
- Ability to Integrate with existing landscape
Based on standard data sets or machine learning prediction capabilities, the accuracy of identifying objects was above 88%