EfficientNet Explained: Rethinking Model Scaling for Convolutional Neural Networks

แชร์
ฝัง
  • เผยแพร่เมื่อ 17 พ.ย. 2024

ความคิดเห็น • 122

  • @nurlubanu
    @nurlubanu 2 ปีที่แล้ว +12

    This is one of the clearest explanations I have seen. You are explaining everything in detail. Thank you!

  • @Aisha182
    @Aisha182 ปีที่แล้ว +1

    It helps me even in 2023 . Hats off to you

  • @zheyeetan4048
    @zheyeetan4048 5 หลายเดือนก่อน +1

    salutation from malaysia. BEST EfficientNet explanation ever!! TQ

  • @_ifly
    @_ifly 8 หลายเดือนก่อน

    It was straightforward and an amazing way of explaining the point.

    • @CodeWithAarohi
      @CodeWithAarohi  8 หลายเดือนก่อน

      Glad it was helpful!

  • @rblsmaa1004
    @rblsmaa1004 19 วันที่ผ่านมา

    This is very helpful . Thank you!

  • @aneerimmco
    @aneerimmco 5 หลายเดือนก่อน

    Clear and non complicated. Thank you.

  • @anshadpa9582
    @anshadpa9582 10 หลายเดือนก่อน

    Thank you for this Well explanation of Compound scaling EfficientNet B0👍

    • @CodeWithAarohi
      @CodeWithAarohi  10 หลายเดือนก่อน +1

      Glad it was helpful!

  • @rupakdey6753
    @rupakdey6753 3 ปีที่แล้ว +1

    Easiest explanation ever . Thank you mam.

  • @ayushkurlekar3302
    @ayushkurlekar3302 2 ปีที่แล้ว

    You are explaining like a pro! Thanks mam!

  • @muhammadsabir6527
    @muhammadsabir6527 ปีที่แล้ว

    wonderfully explained madam. I really liked this video.

  • @Sunil-ez1hx
    @Sunil-ez1hx ปีที่แล้ว

    Very nice & clear video ma’am. Please keep posting👏👏👏

  • @merv893
    @merv893 2 ปีที่แล้ว

    Ok you’ve convinced me EffientNet is the one for me

  • @ahmedshmels8866
    @ahmedshmels8866 2 ปีที่แล้ว

    Wow
    Ma'am amazing explanation with deap knowledge
    No word to say !

  • @takangoudadyavangoudra8761
    @takangoudadyavangoudra8761 2 ปีที่แล้ว

    A great explanation by you, Thank You

  • @usamarajput6418
    @usamarajput6418 4 หลายเดือนก่อน

    this video helped me a lot. thank you Aarohi

    • @CodeWithAarohi
      @CodeWithAarohi  4 หลายเดือนก่อน

      Glad it helped you :)

  • @modernfusions1981
    @modernfusions1981 2 ปีที่แล้ว

    Very good and in detailed explanation. Highly recommended

  • @mdyounusahamed6668
    @mdyounusahamed6668 ปีที่แล้ว

    Great explanation. Thank you.

  • @freya291
    @freya291 2 ปีที่แล้ว +1

    You deserve more views and subscriptions! Thanks a lot for this wonderfull lecture

  • @ariouathanane
    @ariouathanane 2 ปีที่แล้ว

    Thank u very much for this explanations

  • @ShadiHazhir
    @ShadiHazhir 5 หลายเดือนก่อน

    Amazing! I appreciate you, thank you sooo much

    • @CodeWithAarohi
      @CodeWithAarohi  5 หลายเดือนก่อน

      Glad it was helpful!

  • @RanjitSingh-rq1qx
    @RanjitSingh-rq1qx 2 ปีที่แล้ว

    I can't explain in my word. Really mam you are best teacher😊 on the youtub. But i don't know why your subscribers is very less on TH-cam.😔😔, I will share you channel with my friends.👍😊

  • @mehrshad9
    @mehrshad9 ปีที่แล้ว

    Great explanation! keep the hard work! Thank you.

  • @annaperova7092
    @annaperova7092 ปีที่แล้ว

    Thank you! Helped me a lot!

  • @csedepartment236
    @csedepartment236 ปีที่แล้ว

    Thank You Ma'am

  • @YashSharma-le3mo
    @YashSharma-le3mo 10 หลายเดือนก่อน

    Nice explanation mam ❤

  • @digioasis4832
    @digioasis4832 3 ปีที่แล้ว

    Thank you mam, very clear and great explanation.

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว

      You are welcome 😊

    • @digioasis4832
      @digioasis4832 3 ปีที่แล้ว

      Mam when will you post video for grid search.

  • @pramodhbr4190
    @pramodhbr4190 3 ปีที่แล้ว

    Thank you mam. Superb explanation

  • @verma.prashant
    @verma.prashant 3 ปีที่แล้ว +1

    Your videos are very informative but I have suggestion for you that your picture block covers half of ppt, can you reduce your picture block or you can use green screen. Thanks

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว +1

      Thankyou for this suggestion .. Will implement from next video

  • @pramaysimha8789
    @pramaysimha8789 2 ปีที่แล้ว

    Thanks so much for the clear explanation. Could you also please explain Dual Attention Network for segmentation

    • @CodeWithAarohi
      @CodeWithAarohi  2 ปีที่แล้ว

      Will try to cover in my upcoming videos

  • @meghnadeshmukh4524
    @meghnadeshmukh4524 4 หลายเดือนก่อน

    Thank you mam for teaching us sooo nicely.. I totally agree with @shahidulislamzahid... mam you are too good.

    • @CodeWithAarohi
      @CodeWithAarohi  4 หลายเดือนก่อน

      Glad my videos are helping you. keep learning :)

  • @dattatammisetti9324
    @dattatammisetti9324 2 ปีที่แล้ว

    Thankyou Aarohi di 😊

  • @MansaKundrapu
    @MansaKundrapu 2 ปีที่แล้ว

    you have explained it very clearly. can we add skip connection in efficientNet ?

  • @yashrajwani2077
    @yashrajwani2077 2 ปีที่แล้ว

    Thank you. Very helpful

  • @madhurimasarkar5913
    @madhurimasarkar5913 3 ปีที่แล้ว

    It is really good explanation and helpful enough. When implementation of this will be uploaded? Please make a video of DenseNet.

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว +1

      Implementation of EfficientNet will be available in next 2 days. Will do video on DenseNet after finishing my pipelined videos. Keep Watching 😊

  • @spoc.mnmjecspringboardmnmjec
    @spoc.mnmjecspringboardmnmjec 9 หลายเดือนก่อน

    Good mam🎉🎉

  • @mitya7068
    @mitya7068 2 ปีที่แล้ว

    Very useful, thanks a lot!

  • @sajedehtalebi902
    @sajedehtalebi902 ปีที่แล้ว

    great.tnx

  • @sumitchhabra2419
    @sumitchhabra2419 2 ปีที่แล้ว

    Hi,
    Thanks for the amazing explanation.
    I have a question and would request your answer:
    If alpha beta and gamma are fixed. φ is obtained using Grid search. Here I want to understand, that during grid search we have to consider different values of φ.
    Also, I want to understand what is f here. How does the value of f contribute to the Neural Network?
    Please Advise.
    Thank You
    Sumit

  • @YaswanthsriseshagiriNimmakayal
    @YaswanthsriseshagiriNimmakayal 9 หลายเดือนก่อน

    at 32:36 you said the authors already fixed the values of depth width resolution factors, and found phi value, but in that research paper they kept the value of phi as 1 and did a grid search for alpha beta gamma after they found the values they adjusted the phi to upscale different Efficientnets, correct if i am wrong it will be helpful? and the video is good

  • @tiffin2358
    @tiffin2358 8 หลายเดือนก่อน

    Brilliant

  • @samvandenelsaker9576
    @samvandenelsaker9576 2 ปีที่แล้ว

    Thanks a lot!

  • @abhaypratap5311
    @abhaypratap5311 3 ปีที่แล้ว

    Please make a video on 1-D CNN

  • @muhammadsabir6527
    @muhammadsabir6527 ปีที่แล้ว

    Maam do you have accumulative explanation on all the CNN models so that a person can get idea of which CNN model has research gap and needs to be worked on it compare to other cnn models on PhD level research gap.

  • @nau7tico
    @nau7tico 3 ปีที่แล้ว

    Wow, You are awesome. Great explanation. I didn´t understand the last part. How the grid search to obtain (phi) is done?

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว +1

      Will try to cover the grid search in separate video.

    • @santhoshkumarn8585
      @santhoshkumarn8585 3 ปีที่แล้ว

      Grid search is Hyperparameter tuning technique, researcher would have have done investigation here by trying multiple values and finalized the best or optimized values as depth=1.20;width=1.10 and resolution=1.15

  • @clydegriffiths6372
    @clydegriffiths6372 2 ปีที่แล้ว

    Thanks for explanation. I have question. Can we add this ECA to YOLOv5(put somewhere inside YOLO architecture). Can it decrease Loss Function? Guess)

  • @sudharsanb9391
    @sudharsanb9391 3 ปีที่แล้ว

    Mam can u update your playlists by adding the recent videos??so that it will be easy for us to follow.Thank you

  • @rama_gpubhuyan9409
    @rama_gpubhuyan9409 2 ปีที่แล้ว

    Hello,
    I am using pre trained EfficientNetB0 with ImageNet, I have costume 166 image set, I am using regularizations(drop out and batch normalization )/Augmentation technique. But Getting pretty bad result my val_loss increases and val_accuracy remain constant. could you please help.

  • @VipinSingh-he9yc
    @VipinSingh-he9yc ปีที่แล้ว

    Moye moye

  • @fghgffgvbgh
    @fghgffgvbgh 2 ปีที่แล้ว

    Is width scaling means increasing the number of kernels/filters ?

  • @swetajain7929
    @swetajain7929 ปีที่แล้ว

    What is mb6 layer

  • @kritiohri558
    @kritiohri558 3 ปีที่แล้ว +1

    Mam how to calculate phi?

  • @dhanshreepajankar1570
    @dhanshreepajankar1570 2 ปีที่แล้ว

    Hello Ma'am,
    Can we perform Detection of Malware in windows system (output expected is only in YES/NO) using EfficientNet, or is it only meant for image classification?

    • @CodeWithAarohi
      @CodeWithAarohi  2 ปีที่แล้ว +1

      EfficientNet is used Image Classification

  • @sanatshirodkar2161
    @sanatshirodkar2161 2 ปีที่แล้ว

    is it mandatory to use 600*600 dimensions for EfficientNet B7, or can we use smaller dimensions, like say 120*120?

    • @CodeWithAarohi
      @CodeWithAarohi  2 ปีที่แล้ว +1

      The logic behind efficientnet is to work for high resolution images. If you want to wok for small image size you can use efficientnet B0

    • @sanatshirodkar2161
      @sanatshirodkar2161 2 ปีที่แล้ว

      @@CodeWithAarohi Okay, got it! Thank you :)

  • @100_abirhasan8
    @100_abirhasan8 ปีที่แล้ว

    Mam, alpha=1.2,beta=1.1& gamma=1.15 and fie=1 how this value came from f=alpha x beta^fie x gamma^fie. Will you please explain or give relevant resoureces. Thank you.😊😊

    • @CodeWithAarohi
      @CodeWithAarohi  ปีที่แล้ว

      In EfficientNet, the scaling factors for the depth, width, and resolution are denoted as alpha, beta, and gamma. These scaling factors are typically determined through a systematic search on a predefined grid to find the optimal trade-off between accuracy and computational efficiency.

  • @shahrinnakkhatra2857
    @shahrinnakkhatra2857 3 ปีที่แล้ว

    Hi. actually I don't get that part with feature maps, are you referring to channels by feature maps? If yes, then I don't think this is the correct representation for that? (Which you've drawn like bounding boxes)

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว +1

      Hello, Yes I am referring to channels by feature maps. And I understand what do you mean by correct representation : Feature maps should be stacked one after the other. But why I have choose this image, because my intent is to show that different feature maps carry different part of information from image. And if we use more number of feature maps then we can collect more features from image . But again if we choose lots of feature map it will will work upto some extent but after that there is no use of those extra feature maps as those extra feature maps will degrade the training performance. I hope I made my point :)

    • @talha_anwar
      @talha_anwar 3 ปีที่แล้ว +1

      @@CodeWithAarohi i think you are referring to increase number of filters in conv layers

  • @hasanqadhi4280
    @hasanqadhi4280 ปีที่แล้ว

    ❤❤

  • @daily-technology7269
    @daily-technology7269 3 ปีที่แล้ว

    Reduce the size of your own window (at the bottom left), will make the PowerPoint presentation more visible. Otherwise its super annoying when your window is cutting off the text and you have not shared any other document as well. Nobody is here to see some enlarged version of the white space in your wall.

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว

      Thankyou for the feedback.

    • @pawanbhandarkar7647
      @pawanbhandarkar7647 2 ปีที่แล้ว

      This is the WRONG way of giving feedback.
      Alternatively, you could have said:
      Great video! I appreciate the content you have provided. However, I suggest that you reduce the size of your video inset a bit so that it doesn't overlap with the powerpoint presentation, as it sometimes hides the text behind it. Additionally, it would be helpful if you share some documents (like the presentation you used in this video) so that we can refer to the unclear parts.
      Thanks

  • @vinothsomasundaram9519
    @vinothsomasundaram9519 3 ปีที่แล้ว

    Hi i am doing research in resnet in thermal image for face recognition , can you help me?

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว

      Hello, mail me your requirement at aarohisingla1987@gmail.com Will see if I can help you

  • @aftabshaikh5352
    @aftabshaikh5352 3 ปีที่แล้ว +1

    Mam, with all due respect. Please don't repeat too much !!!

    • @CodeWithAarohi
      @CodeWithAarohi  3 ปีที่แล้ว

      Thankyou for the feedback and Yes I am already working on this.

  • @md.alamintalukder3261
    @md.alamintalukder3261 2 ปีที่แล้ว

    Thank u so much

  • @shahidulislamzahid
    @shahidulislamzahid 5 หลายเดือนก่อน

    This is one of the clearest explanations I have seen. You are explaining everything in detail. Thank you!

    • @CodeWithAarohi
      @CodeWithAarohi  5 หลายเดือนก่อน +1

      Glad it was helpful!