laibert

Thanks! Haven't seen that paper, I'll check it out. I think quantization only helps with inference speed if the network is running on CPU with negligible gains on GPU (Tensorflow only supported CPU on mobile last I looked which was a while ago). However your app is already super fast so don't I think anyone would notice if it was marginally faster at this point!

This is amazing - impressed by your persistence to source the training data yourself, that must have been tedious!

Did you try quantizing the parameters to shrink the model size some more? If so, how did it affect the results? It also runs slightly faster on mobile from my experience.

This is the setup I recently made and has been working great so far. However, most of it was purchased on black friday - you can probably replace any of these components with comparable parts that are currently on sale.

https://docs.google.com/spreadsheets/d/1GgTpPbLdRDvw3Mf_d1W0...

True. I'm not trying to diminish the importance of Apple hardware, but the design philosophy of the hardware is minimal because the software is the focal point. In the end, the phone is a rounded rectangle of varying degrees of thinness. IOS is the reason why the iPhone is a joy to use: the metal does its best job to stay out the way. Responsive touch screens and awesome cameras allows the user to forget they're even using a phone.

While my personal opinion regarding skeumorphic design falls very much in line with the author's, I don't think that there is a contradiction in Apple's design philosophy.

The hardware is minimal because its true purpose is to let the software shine. It's a magical box that can become anything, from a calendar to a photobooth to a music studio. Apple's visual design choices in software have always seemed to fall under the skeuomorphic end of things (faux brushed aluminum?) and having a minimal hardware accentuates these flourishes.

I'm not sure I understand how opening up classes to the online world would would affect the coursework of Stanford students. If the author felt that CS229a was not rigorous enough, he should have just taken CS229 whose problem sets could double as full-time jobs.

Morever, a good chunk of stanford CS classes are already offered online through SCPD (the professional development system) and as far as I know, the structure of those classes remain unchanged.

are you on High and Forest?!