TensorFlow.js: From prototype to production, what's new in 2022?

[@jhonshephard921]

I just had a thought about the framerate. If you webcam is recording at 60 or less fps, and then its feeding the frames to the model, would the webcam become the bottleneck in the inference framerate? because you have the recording rate, then the time it takes to go to the model, and then finally the time it takes the model to give you an output.

[@JasonMayes]

Great question, so yes if your webcam only did 30 fps or 60 like most do, then if your model is running 120fps then it will simply be classifying the same image 2 (60fps) or 4 (30fps) times! If you know this to be the case you can limit how often you call the ML model so you don't waste processing power on the same thing

[@hamdaghribi111]

@@JasonMayes qdfrrr 0r4rer4411ppp000v0prerwwwwwwwwwwghy6776gexvbv1aa

[@JasonMayes]

If you have any questions do reach out to me on Twitter (jason_mayes) or LinkedIn (search my name).

[@jhonshephard921]

Hi, great video. I don't see the course you mentioned in the description. It just links to the main tensorflow website.

[@JasonMayes]

@@jhonshephard921 the link in the video works for me. If issue go to edx website edx dot org and search TensorFlow JS there is only one course by me there!

[@ozzietradie6514]

Absolutely amazing technology 👏 exactly what I was looking for.The team at Google should be proud of this work.

[@JasonMayes]

@@ozzietradie6514 thank you for the kind words! Look forward to seeing what you make with it! Do tag me if you do

[@umsief178]

👍👍👍

[@basheershahrour593]

Great !

[@innthomas]

link to the tfjs course

[@JasonMayes]

goo.gle/learn-tfjs

[@jasonmayes_goog]

goo.gle/learn-tfjs