With the increasing prevalence of automated and assisted driving, the demand for edge computing is also on the rise.
How to assist customers in developing AI applications has become a new topic. Last year, WPG introduced the eIQ edge computing solution, for more details please refer to: here. In 2020, we are introducing the new eIQ 2.0, which significantly enhances efficiency and various usage scenarios.
This solution is developed based on NXP’s native BSP 5.4.24_2.1.0, incorporating Python elements, and can utilize GPU/NPU for AI neural network computations, making both efficiency and application scenarios more comprehensive.
First, install the Python package using pip:
The image recognition example uses the classic image grace_hopper.bmp: In the early version of eIQ in 2019, the inference time was about 330ms. In 2020 with eIQ 2.0 (PyeIQ):
It can be seen that with GPU computation capabilities, using the same mobilenet model, TensorFlowLite only requires about 10ms, achieving a speed and efficiency improvement of 30 times!! Additionally, eIQ 2.0 provides real-time image input and video recognition capabilities:
Using videos shot on real streets in Taiwan for object recognition projects shows that the actual performance of the i.MX8QM is 30ms. This solution provides the following NNAPI tables for development across various platforms and algorithms:
This solution integrates various algorithms and provides corresponding APIs for customer development. Coupled with the i.MX8QM’s 4-core A53 + 2-core A72 + 2 GC7000XSVX GPU, it can utilize GPU computation for AI and image technology while stably providing system resources. Several customers have already developed automotive market applications using this solution. The WPG FAE team will use this solution as a foundation to assist all customers in developing applications in the AI field.
► Application Scenario Diagram
► Solution Block Diagram
► Core Technical Advantages
-
Automotive Grade, ASIL-B
-
16x Vec4-Shader GPU, 32 compute units OpenGL® ES 3.2 and Vulkan® support Tessellation and Geometry Shading
-
2x ARM A72 core + 4 A53 core
-
MIPI CSI can connect two HD cameras simultaneously
-
WPG provides cross-platform (PC to I.MX) ML (Machine Learning) applications
► Solution Specifications
Python 3.7
TensorFlow 2.1
TensorFlowLite 2.1
OpenCV 4.2.0
ArmNN 19.08
Click to read the original text, download the technical documentation of the solution, unlock over 1000 IoT, automotive, and power solutions, and connect with over 700 technical experts online. Click “See” to give the editor a thumbs up