Inception: GoogLeNet Architecture (in Hindi)

[@prateeksahu8529]

sir you are so good in teaching, you made for teaching .

[@6kunal9]

This channel is amazing! Great job sir, you are doing a Fabolous job .

[@KshitizVermaDL]

Thank you very much!

[@priyanshkhunger1283]

awesome explanation, you gained a new subscriber

[@KshitizVermaDL]

Thank you!

[@pallabidas3064]

Nice video sir. Thanks a lot🙏

[@KshitizVermaDL]

You are most welcome.

[@technopathkeshav7835]

Amazing...😊😊

[@KshitizVermaDL]

Thanks

[@methilmuley9047]

Sir, what I have to do if I have 224*224*1 dimensional image and I have to use googlenet.

[@KshitizVermaDL]

Have you been through the basics of CNNs? They should help you.

[@aishwaryavpkadam]

what did you do?

[@moomalpanwar9947]

hi can you please help me find the number of layers in my inception network made for mtech final project. I am confused how to calculate the number of layers.

[@KshitizVermaDL]

Hi Moomal. Thanks for the comment. However, it is difficult for me to reach out to individual requests. Sorry!

[@manishswami877]

thank you sir

[@amandhamanda9822]

Sir complexity nhi bdegi like itne sare filters...

[@KshitizVermaDL]

Bilkul badhegi!! But authors work at Google! They have the highest quality compute resources. They could manage the increased complexity. In those days, or may be even today, if you are able to train a huge number of parameters successfully, most likely you will have a paper in a tier 1 machine learning conference.

[@amandhamanda9822]

@@KshitizVermaDL sir axha samja...sir eek question har architecture apne m unique h.... So sir koi guidance kb konsa follow kre...like waise toh resnet vgg k compare m jyada axha h...but like resnet aur inception ko bhi compare kre kaise pta chlega konsa use krna h...like inception i think more axhi performance dekega mere understanding se...isko sir elaborate krke btana

[@KshitizVermaDL]

All the architectures have their pros and cons. You have to see which one fits your task. If you are training on your data, taking inspiration from such networks is good but you may have to design your own architecture. If you don't have compute, use transfer learning. See the video on transfer learning. I think that should help you answer your question up to some extent.

[@amandhamanda9822]

@@KshitizVermaDL okay...Thanks sir

[@jayantapaulofficial]

sir, what is training error and test error?

[@KshitizVermaDL]

You need to go through the earlier videos. Such things are discussed in detail. See videos in playlist on regularization.