AEs seem overkill for this problem. There are plenty of shallow-learning (non-ml) DSP filters which can do this with lower overhead on non-custom hardware.
Thanks for watching! This is one simple demo that we showed at a conference, but there are countless applications that benefit from small-scale, real-time neural networks. Check out some other examples: apis.liquidinstruments.com/mnn/
@@LiquidInstruments Would be nice if people would stop using the nomenclature of "neural networks" when at most, this is machinelearning. The term "neural network" was deviced by McCulloch & Pitts back in the 1930's. Their oversimplification has nothing to do with a neural network whatsoever. Same goes for the other overused buzzword, "AI" which again, is at most just machinelearning. Sorry for the rant, I just find it misleading using these types of buzzwords. The product otherwise seems interresting.
@@Gin-toki A neural network is not a buzzword, it is something very concrete. An autoencoder is a neural architecture, that is, a kind of neural network. Yes, pre-trained neural nets are used here to denoise signals in real time. You can look up the term "denoising autoencoder" to verify.
I need this to implement on acoustic signals from AUV's pinger, how do I get started?
you should probably contact them
We're here to help! Reach out to us at the link below to chat with an engineer:
liquidinstruments.com/support/contact-us-general-sales/
AEs seem overkill for this problem. There are plenty of shallow-learning (non-ml) DSP filters which can do this with lower overhead on non-custom hardware.
Thanks for watching! This is one simple demo that we showed at a conference, but there are countless applications that benefit from small-scale, real-time neural networks. Check out some other examples:
apis.liquidinstruments.com/mnn/
@@LiquidInstruments Would be nice if people would stop using the nomenclature of "neural networks" when at most, this is machinelearning. The term "neural network" was deviced by McCulloch & Pitts back in the 1930's. Their oversimplification has nothing to do with a neural network whatsoever. Same goes for the other overused buzzword, "AI" which again, is at most just machinelearning.
Sorry for the rant, I just find it misleading using these types of buzzwords.
The product otherwise seems interresting.
@@Gin-toki A neural network is not a buzzword, it is something very concrete. An autoencoder is a neural architecture, that is, a kind of neural network. Yes, pre-trained neural nets are used here to denoise signals in real time. You can look up the term "denoising autoencoder" to verify.
Lego Batman gif