Introduction to TengineKit
TengineKit, developed by OPEN AI LAB, is an innovation in the realm of AI algorithm software development kits. It's tailored to be easily integrated and is optimized for low latency performance across a wide range of mobile devices. The team at OPEN AI LAB is committed to continually updating TengineKit, ensuring its efficacy and performance remain at the cutting edge of technology.
Key Features
TengineKit shines with an impressive suite of features dedicated to facial, body, and hand detection:
-
Face Detection and Landmark Identification: TengineKit is capable of detecting faces and identifying key landmarks, both in two-dimensional and three-dimensional spaces.
-
Iris and Eye Landmark Detection: The SDK offers precise iris and eye landmark recognition.
-
Body Detection and Landmark Identification: The software can effectively detect upper body parts and their landmarks, which is particularly useful for interactive applications.
-
Hand Detection and Landmark Recognition: While real-time hand detection and landmark recognition are in development for mobile platforms, TengineKit supports these features in other environments.
-
Face Attributes: The software can discern facial attributes such as age, gender, smile detection, and the presence of glasses.
-
Integration with YOLOv5: TengineKit also supports the integration of YOLOv5, a popular object detection model.
Application and Accessibility
TengineKit provides an accessible demonstration of its capabilities through an APK, which can be directly downloaded and installed on Android devices. Users interested in exploring more can scan a QR code to easily obtain this application.
Performance
TengineKit demonstrates exceptional performance across various smartphone models:
- On a Kirin 980, the kit executes face detection and landmark identification tasks in just 4ms, achieving a rate of 250 frames per second (fps).
- Qualcomm 855 processes similar tasks in 5ms, with a performance capacity of 200 fps.
- Older or less powerful models like the Qualcomm 450B require 18ms, still reaching a respectable 56 fps.
Development Goals
The core objectives of TengineKit are ambitious yet focused. The development team aims to deliver:
- Superior Performance: Ensuring that the software offers the best performance on mobile devices.
- Simplicity: Maintaining an easy-to-use API for seamless integration.
- Efficiency: Developing the software with the smallest possible package size for mobile clients.
Continued Development and Updates
As of March 25, 2021, TengineKit received updates enhancing its performance and offerings:
- Fixes for Linux sample code.
- Improvements to Android sample code, increasing frames per second.
- Updates to Linux shared objects and YOLOv5 models.
- Memory optimizations within the core, v0.0.6.
Community and Support
For those interested in learning more about TengineKit or engaging in discussions about facial recognition technologies, a QQ group has been established. This provides a platform for technical exchange and support among users and developers.
In summary, TengineKit stands out as a robust toolkit for AI algorithm integration in mobile environments. It boasts a host of features catering to facial, body, and hand detection, with a commitment to continuous improvement and community engagement.