[](LICENSE)

CaffeOnACL is a project that is maintained by **OPEN** AI LAB, it uses Arm Compute Library (NEON+GPU) to speed up [Caffe](http://caffe.berkeleyvision.org/) and provide utilities to debug, profile and tune application performance.

* The ARM Computer Vision and Machine Learning library is a set of functions optimised for both ARM CPUs and GPUs using SIMD technologies. See also [Arm Compute Library](https://github.com/ARM-software/ComputeLibrary).

* Caffe is a fast open framework for deep learning. See also [Caffe](https://github.com/BVLC/caffe).

* [Installation instructions](https://github.com/OAID/CaffeOnACL/blob/master/acl_openailab/installation.md)

* [User Manuals PDF](https://github.com/OAID/CaffeOnACL/blob/master/acl_openailab/user_manual.pdf)

* [Performance Report PDF](https://github.com/OAID/CaffeOnACL/blob/master/acl_openailab/performance_report.pdf)

There are some compatibility issues between ACL and Caffe Layers, we bypass it to Caffe's original layer class as the workaround solution for the below issues

* Normalization in-channel issue

Performance need be fine turned in the future

The Caffe based version is [793bd96351749cb8df16f1581baf3e7d8036ac37](https://github.com/BVLC/caffe/tree/793bd96351749cb8df16f1581baf3e7d8036ac37).

Support Arm Compute Library version 17.06 with 4 new layers added

* Batch Normalization Layer

* Concatenate layer

Fix the issues:

* Compatible with Arm Compute Library version 17.06

* When OpenCL initialization fails, even if Caffe uses CPU-mode,it doesn't work properly.

Initial version supports 10 Layers accelerated by Arm Compute Library version 17.05 :