Released ailia SDK 1.2.11

We are pleased to introduce version 1.2.11 of ailia SDK, a cross-platform framework to perform fast AI inference on GPU or CPU. You can find more information about ailia SDK on the official website.

Support for new operators

The support for operators BitShift、CumSum、GatherElement、ThresholdedRelu was added, along with the Sequence input for Loop.

Faster loading of obfuscated models

The use of AES-NI speeds up the loading of obfuscated models for Windows (Intel), macOS (Intel), Linux (Intel), Android (Arm64) and iOS (Arm64) environments. For example, a 1.24 GB encrypted model can now be loaded 65 times faster.

Reduced memory consumption of Vulkan

When performing convolution with a large number of channels in Vulkan, memory consumption is reduced by dividing the work memory.

cuDNN8.3 support

A problem with ailiaCreate returning an error when using ailia SDK with cuDNN8.3 was fixed.

Unity plugin

ailia’s Unity plugin has been upgraded to support Unity 2019.4.32f1 and later. In addition, we have fixed a problem that caused excessive error logs to be displayed when the camera was not available or when the license file did not exist.

Change Python Environment props to list

Python Environment props is now given as listinstead of str to support multiple flags. No changes to the source code are necessary if you are using the in operator as it is done in ailia-models samples. The flag FP16 is now given to props when using macOS MPS.

Evaluation version

Starting with version 1.2.11, the license file is now required for RaspberryPi and Jetson platforms. The license file must be placed in the ~/.shalo/ folder.

---

ailia Inc. has developed ailia SDK, which enables cross-platform, GPU-based rapid inference.

ailia Inc. provides a wide range of services from consulting and model creation, to the development of AI-based applications and SDKs. Feel free to contact us for any inquiry.