this is really cool compilation of what ive been doing in the comp vision course im completing.. still feels like for me personally id have to look at each of these a little more in depth to fully grasp but still great video! Maybe a suggestion would be a " less edited" real time maybe live of how you tackle going through these papers and maybe explaining underlying math concepts?
Thanks for the feedback! The video was definitely intended to introduce the major concepts in the field and give direction/references for further self study (and not explain each topic elaborately!)
Excelent video! i have a doubt, i have read that LeCun and his team isn't that they created CNN's, instead they improved it and applied it, the actually creator was fukushima, im wrong?
Thanks, glad you enjoyed the video. And I think you are correct! Thanks for pointing it out. You must be referring to the Neocognitron architecture (www.rctn.org/bruno/public/papers/Fukushima1980.pdf) , which can be thought of as precursors to CNNs and iirc I think LeCun references it in his paper as well. LeCun and his team developed the modern version of CNNs and trained it with backprop (which earlier models like the Neocognitron didn't).
Understandable! Lot of complex concepts going on here but I didn’t want the video to go over 30 minutes so I tried to explain as precisely as I could. I thought people can slow down or pause or rewatch sections that they found tricky. There’s also a medium article here that I wrote if it helps: medium.com/@neural.avb/the-history-of-convolutional-neural-networks-for-image-classification-1989-today-5ea8a5c5fe20
Hey, thanks! I generally use a combination of matplotlib (python), Manim, Powerpoint, Cavalry (it's a free vector graphics software), and Davinci Resolve (video editing software) to produce my videos.
This video has taught me more in 20 mins than I've learned in the entire last week. Wonderful explanation, thank you very much!
That was cool thanks for making this video would be cool to implement some and explore how can be used in all kind of cool ways and explore
Thanks mam you have cleared my many doubts.
Concise summary!! The animations look great.
solid
this is really cool compilation of what ive been doing in the comp vision course im completing.. still feels like for me personally id have to look at each of these a little more in depth to fully grasp but still great video! Maybe a suggestion would be a " less edited" real time maybe live of how you tackle going through these papers and maybe explaining underlying math concepts?
Thanks for the feedback! The video was definitely intended to introduce the major concepts in the field and give direction/references for further self study (and not explain each topic elaborately!)
goat
Loving this video. Cheers!
Thank you! Cheers!
Your videos are so good for people actually in industry. I really appreciate everything you're putting out!
Glad that finally someone made this video! Thanks a lot. Best rundown of CNNs on TH-cam.
amazing, I allways wondered where all these pretrained models came from
Really great and sleek explanation.
Excelent video! i have a doubt, i have read that LeCun and his team isn't that they created CNN's, instead they improved it and applied it, the actually creator was fukushima, im wrong?
Thanks, glad you enjoyed the video. And I think you are correct! Thanks for pointing it out. You must be referring to the Neocognitron architecture (www.rctn.org/bruno/public/papers/Fukushima1980.pdf) , which can be thought of as precursors to CNNs and iirc I think LeCun references it in his paper as well. LeCun and his team developed the modern version of CNNs and trained it with backprop (which earlier models like the Neocognitron didn't).
Great idea and excellent execution. Can you do the same for audio and speech? There are limited resources in this area.
Great suggestion! I’ll look into it.
A fascinating video. But you were a bit fast.
Understandable! Lot of complex concepts going on here but I didn’t want the video to go over 30 minutes so I tried to explain as precisely as I could. I thought people can slow down or pause or rewatch sections that they found tricky. There’s also a medium article here that I wrote if it helps:
medium.com/@neural.avb/the-history-of-convolutional-neural-networks-for-image-classification-1989-today-5ea8a5c5fe20
Hey this was fantastic btw which software you are using to make these , keep doing the good work
Hey, thanks! I generally use a combination of matplotlib (python), Manim, Powerpoint, Cavalry (it's a free vector graphics software), and Davinci Resolve (video editing software) to produce my videos.
@@avb_fj holy moly it's a lot of softwares