Multiple Camera AI on AM62A

Demo Background

With the advances of deep learning models, more applications utilize multiple camera inputs which provide tremendous benefits which are not possible with a single camera. Some of the applications which can benefit from multi-camera inputs include surround view, cabin recorder and camera mirror system, medical imaging, security surveillance and many more.

AM62A is equipped with a deep learning accelerator (C7x-MMA) with up to 2 TOPS which is capable of running various types of deep learning models for classification, object detection, or semantic segmentation simultaneously. This demo presents a reference design of running multiple-camera applications on AM62A EVM, using the Arducam V3Link Camera Solution Kit (https://docs.arducam.com/V3Link-Camera-Solution/V3Link-Camera-Solution-on-TI-Platform/Introduction/) to connect up to 4 CSI-2 cameras to AM62A and running deep learning models for all 4 cameras. This V3Link kit works with both FPD-Link and V3-Link based cameras.

The he four cameras AI demo operated at 30 FPS each with total 120 FPS for the four cameras. Even with this high rate, only 86% of the deep learning accelerator C7x-MMA is utilized. This rate is limited by the performance of the cameras used in the testing (imx219) which does not exceed 30 FPS. In addition, the deep learning accelerator is clocked at 850 MHz instead of 1000 MHz. This means that the in these experiments, the C7x-MMA is running at about only 85% of its expected performance. This extra bandwidth of the deep learning accelerator can be utilized for bigger deep learning models or higher frame rate depending on the targeted application.

How to get started

Following are the steps to run the demo:

1. Get an AM62A starter kit EVM
2. Get an Arducam V3Link Camera Solution Kit
3. Download the Edge AI Linux SDK
4. Load the Edge AI Linux SDK via an SD card using the quick start guide
5. Log into the EVM through a network connection or a serial port connection
6. Follow the steps in Developing Multiple-Camera Applications on AM6x app note.

Additional Resources for the Multi Camera Demo Demo

General Edge AI and AM62A resources

Please find the following resources related to the AM62A and TI Edge AI.