Mobile Depth Learning mobile-deep-learning (MDL)

Free and open source mobile deep The learning framework, deploying by Baidu.

This is the simply deploying CNN on mobile devices with the low complexity and the high speed. It supports calculation on the IOS GPU, and is already adopted by the Baidu APP.

size:340k+ (on ARM v7)
Speed:40ms (for IOS Metal GPU mobilenet) or MS (for Squeezenet)
Baidu Research and development of the mobile end of the deep learning framework, is committed to let the convolution neural network extremely simple deployment on the mobile side. is currently operating in mobile phone Baidu. Support for iOS GPU computing. Small size, fast speed.

Volume ARMv7 340k+
Speed IOS GPU mobilenet can reach 40ms, squeezenet can reach 30ms

Project Address: https://github.com/baidu/mobile-deep-learning

More Machine learning Tutorials: http://www.tensorflownews.com

Characteristics

One-click Deployment, script parameters can switch iOS or Android
support iOS GPU run mobilenet, squeezenet model
has been tested to run stably mobilenet, Googlenet v1, The Squeezenet model
is small in size and without any third party dependencies. Pure hand-built.
provides quantitative scripting, direct support for 32-bit float to 8-bit uint, model volume quantization 4M up and down
and arm-related algorithm team line under the online multiple communication, for the arm platform will continue to optimize the
Neon use covers convolution, normalization, pooling all aspects of the operation
assembly optimization, specific optimization for register assembly operations
Loop unrolling loop expansion, reducing unnecessary CPU consumption for performance improvement, all expand judgment operations
Place a large number of heavy computing tasks in front of the overhead process