Featured Articles
Announcing Pi
Technical Specifications
Raspberry Pi FAQs
Win a Pi
Raspberry Pi Wishlist
Actions
Forum Thread Details
  • Replies 12 replies
  • Subscribers 674 subscribers
  • Views 286 views
  • Users 0 members are here
Related

Looking for ideas: cheap but awesome project ideas

micheal.embedded

Hi everyone,

I’m currently in my final year at university and I’m required to design and build a project for my final-year submission. I’m really excited about this because I want to create something meaningful and technically solid.

The challenge I’m facing is budget  I can’t afford expensive components or complex lab equipment. Because of that, I’m looking for ideas on how to build something that is low-cost but still innovative and impressive from an engineering point of view.

Parents

  • What counts as low-cost? How complex are we talking? Is the focus mostly on software, or both hardware and software?

    I've been looking at the NUCLEO-N657X0-Q, it has an insanely fast core and a NN accelerator. You mentioned embedded systems and AI related, so maybe combine the two?
    Make and train some relatively small model for a task. Think object detection, voice recognition, or even something like playing chess. 

    If I'd have to pick something now, I'd probably get some camera module for it, and train a network to recognize chess pieces on a board, and another one to play chess(although pure NNs are not the optimal way to do it, look into how Stockfish works if you're interested). See if the NN accelerator is powerful enough to run some existing pretrained voice synthesis model. Recognize chess pieces, figure out the best move, say the move out loud. A little portable chess machine that works with regular chess boards.

Reply

  • What counts as low-cost? How complex are we talking? Is the focus mostly on software, or both hardware and software?

    I've been looking at the NUCLEO-N657X0-Q, it has an insanely fast core and a NN accelerator. You mentioned embedded systems and AI related, so maybe combine the two?
    Make and train some relatively small model for a task. Think object detection, voice recognition, or even something like playing chess. 

    If I'd have to pick something now, I'd probably get some camera module for it, and train a network to recognize chess pieces on a board, and another one to play chess(although pure NNs are not the optimal way to do it, look into how Stockfish works if you're interested). See if the NN accelerator is powerful enough to run some existing pretrained voice synthesis model. Recognize chess pieces, figure out the best move, say the move out loud. A little portable chess machine that works with regular chess boards.

Children
No Data