How Much Hardware Does Edge AI Really Need?

https://hackster.imgix.net/uploads/attachments/2007153/_VEhrIwqM6I.blob?auto=compress%2Cformat&w=600&h=450&fit=min

GreatScott’s autonomous lawn mower can navigate with centimeter-level precision using RTK GPS, but that accuracy comes at a price. After a viewer suggested replacing the expensive positioning hardware with a camera and AI, he decided to put the idea to the test. Instead of knowing exactly where the mower is, a vision system could simply look at the ground, recognize where the grass ends, and tell the robot when it needs to turn.

That raised another question: How much computer does something like that actually require? To find out, GreatScott compared three edge AI platforms ranging from a $20 microcontroller setup to a $200 single-board computer equipped with dedicated AI hardware. Each was tasked with running computer vision models that could distinguish grass from things the mower should avoid, such as gravel, stone, or wooden walkways.

A Sipeed MaixCAM (📷: GreatScott)

The cheapest option is the Seeed Studio XIAO ESP32S3...

Copyright of this story solely belongs to www.hackster.io. To see the full text click HERE

Read more

https://techcrunch.com/wp-content/uploads/2022/07/CMC_1580.jpg?resize=1200,800

Apple says it is adding additional controls around “Full Disk Access” on macOS as AI agents have increased “the risks associated with this level of access”

Sponsor Posts Subquadratic: the LLM built for 12M-token reasoning — SubQ can reason across entire codebases and document sets in one pass with no RAG workarounds. Read how SubQ 1.1 Small holds near-perfect retrieval out to 12M tokens. Introducing Campus: The digital home for educational institutions — Every educational institution needs