2024 BrainChip Discussion, page-7235

The first link is to Prophesee’s website and includes a paper from 2020 identifying the need for a event based data set for training event based systems for assisted driving and goes on to set out what is needed to process that data and concludes the paper by saying the problems identified would be solved by using event based hardware. They provide two references one to IBM’s True North and one to Intel Loihi.

We know that before Prophesee said that Brainchip AKIDA had completed their house of straw they trialed chips from Intel and SynSense.

https://www.prophesee.ai/2024/07/16/learning-to-detect-objects-with-a-1-megapixel-event-camera/

The second paper is from Brainchip published on 31.5.24 which does two things relevant to the above paper. It solves all the problems Prophesee identifies in 2020 and then runs Brainchip’s solution on the Prophesee data set with state of the art performance:

https://arxiv.org/pdf/2405.12179

From my point of view my interest was triggered by the comment of Chris Stevens in his video presentation that AKIDA 2.0 TENNS overcame the problem with event based sensors not seeing moving objects that became stationary by using memory.

When you read the Prophesee paper they speak about this problem.

There is a bit of reading but you have the weekend ahead so enjoy.

My opinion only DYOR

Fact Finder