@@ShriramVasudevan It looks there is small mistake at 2:06 . I guess we will have only 2 fc with 4096 and one softmax. But in the slide there are 3 fc 4096.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
@Shriram Vasudevan It looks there is small mistake at 2:06 . I guess we will have only 2 fc with 4096 and one softmax. But in the slide there are 3 fc 4096.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Great video. I have only one question. What are the dimensions of your training and test images? Do they have to be 224*224 before you import them into the algorithm?
Yup! Training and Validation images should be 224x224 because vgg16 was trained with 224x224 images (ImageNet ) ! But they do not have to be 224x224 before feeding them to the CNN, you can resize them while augmenting the images with: img_generator = imgDirectory.flow_from_directory(target_size=(224, 224), ... )
hi sir, can you please explain what are input rgb channels doing in conv2d ,cause i can see 1st layer output channels are 64 that means we consider black and white image?
Hi, Shriram thank you for your knowledge sharing. I have couple of doubts you printed model summary before building or fitting the model, secondly when you execute this error pops up:ValueError: Shapes (None, 1) and (None, 2) are incompatible for categorial_cross entropy loss and softmax in the last layer. can you give idea how you approached to resolve these two in your demo code shown in this video.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Thanks for delivering such an informative lecture. You have written if output[0][0] > output[0][1]: print("cat") else: print("dog") How to write if-else loop of 5 classes?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hi Sir, Why use Conv layers with the same parameters in a block. ex the first block Conv64 x 2 or the fifth block Cov512 x 3. Thank you very much! ( my listening skill is not good)
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hi Shiram - When I try to execute the code in google colab, I got an exception (InvalidArgumentError: logits and labels must be broadcastable) for this line : hist = model.fit_generator(steps_per_epoch=100,generator=traindata,validation_data=testdata,validation_steps=10,epochs=5). Have you come across this kind of issues?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Error = logits and labels must be broadcastable: logits_size=[32,2] labels_size=[32,3] Solution = you are having just two classes in your final classification layer while you actually need 3
Sir I wanted to know that you have approx. 21 lacks parameters among which only approx. 6 lakhs are trainable and rest of them are non trainable ..So Does it will matter for good CNN Model. Actually I wanted to build my own CNN Model for my research work
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hey Shriram, thanks a lot for the video! But I think there is some mistake, because the last layer should return 1 unit and not 2. It didn't run in my colab at all, until I changed the last dense later to 1 from 2. I was getting some error from keras because of that.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hi Sir, its not necessary that all the input images will be of size 224 * 224, some of the images which I have downloaded from google are far less than the said size, how to handle such images. Please guide.
YOU can reshape the size of all the images in a folder, with the various preprocessing functions, and all the images in a folder will be resized to desired size and then deed it.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
python . that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
I get this error : OSError: Unable to open file (unable to open file: name = 'vggweights.h5', errno = 2, error message = 'No such file or directory', flags = 0, o_flags = 0) How do I find out the path location to the downloaded VGG code?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
@@sagarprithvipalsinghrawat6655 We are only using top 13 layers for our model from VGG16. Classification part(After flattening) is upto us. We can use any number of neurons
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
This is perhaps the easiest of all the approaches available online. Good work.
This is the most lucid explanation about vgg16 implementation.Thank u so much Sir
Thanks and glad you liked.
This is the first lecture that tells me how to implement a VGG-based classifier. Great .
Thanks and glad you liked it
@@ShriramVasudevan It looks there is small mistake at 2:06 . I guess we will have only 2 fc with 4096 and one softmax. But in the slide there are 3 fc 4096.
no mine is the first actually implementing
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
very well explained sir ... i saw many videos on the same topic but your way of explaining it was very easy and the best.
Keep watching
Well done, man, it's a "pocketbook" reference even on 2024, be proud!
Thank you brother
@@ShriramVasudevan , it's, absolutely, true. Keep rocking!
@@fredericofurlan :)
one of the best teacher here. thank you sir.
Thanks Isha
@@ShriramVasudevan sir I'm not able to find the pre trained model weights file "vggweights.h5". pls help me with that
Sir, your explanation is outstanding and simple
Thanks and god's grace
Sir, hats off to you for this clear n crisp implementation💯💯.
Glad u liked it Farhan
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
@Shriram Vasudevan It looks there is small mistake at 2:06 . I guess we will have only 2 fc with 4096 and one softmax. But in the slide there are 3 fc 4096.
Thank you sir!! Your explanation was very easy to understand.
Thanks swati
Liked before watching.. undoubtedly the video must be great..
Thanks
Thank you for sharing your valuable knowledge
Thanks
I like the "pooing" layers. I believe that's unique.
Yes.
everything is indeed unique, that's what makes everything work in combination.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Great video. I have only one question. What are the dimensions of your training and test images? Do they have to be 224*224 before you import them into the algorithm?
Yup! Training and Validation images should be 224x224 because vgg16 was trained with 224x224 images (ImageNet ) ! But they do not have to be 224x224 before feeding them to the CNN, you can resize them while augmenting the images with: img_generator = imgDirectory.flow_from_directory(target_size=(224, 224), ... )
hi sir, can you please explain what are input rgb channels doing in conv2d ,cause i can see 1st layer output channels are 64 that means we consider black and white image?
Hi, Shriram thank you for your knowledge sharing. I have couple of doubts you printed model summary before building or fitting the model, secondly when you execute this error pops up:ValueError: Shapes (None, 1) and (None, 2) are incompatible for categorial_cross entropy loss and softmax in the last layer. can you give idea how you approached to resolve these two in your demo code shown in this video.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Why you change 4096 become 256, in the Fc ?
sir why 512 is retained??
Thanks for delivering such an informative lecture.
You have written
if output[0][0] > output[0][1]:
print("cat")
else:
print("dog")
How to write if-else loop of 5 classes?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
sir, where do i get the dataset file?
thanks for this amazing representation.
Thank you brother
sir please reply in that whether we can use 'image net ' for weights instead of vggweights.h5??
Yes
Very nice explanation.
Thanks.
Great lecture sir...Is this what is known as VGG 16 with transfer learning?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hi Sir, Why use Conv layers with the same parameters in a block. ex the first block Conv64 x 2 or the fifth block Cov512 x 3.
Thank you very much!
( my listening skill is not good)
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hi Shiram - When I try to execute the code in google colab, I got an exception (InvalidArgumentError: logits and labels must be broadcastable) for this line : hist = model.fit_generator(steps_per_epoch=100,generator=traindata,validation_data=testdata,validation_steps=10,epochs=5). Have you come across this kind of issues?
Let me have a look brother..
@@ShriramVasudevan Did you get time to look into the issue?
@@yashwanthbkumar1755 did find solution?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Error = logits and labels must be broadcastable: logits_size=[32,2] labels_size=[32,3]
Solution = you are having just two classes in your final classification layer while you actually need 3
Great lecture .. Thank you.. can you please share these slides ,
These slides are part of my book. So it's difficult.
How to change the size of training dataset with size 28,28,3 to VGG16 compatible input_shape
Sir I wanted to know that you have approx. 21 lacks parameters among which only approx. 6 lakhs are trainable and rest of them are non trainable ..So Does it will matter for good CNN Model. Actually I wanted to build my own CNN Model for my research work
Simple and superb
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
On minute 11, if i have 6 different classes, how i want to add another classes such as rabbit, alligator etc?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hey Shriram, thanks a lot for the video! But I think there is some mistake, because the last layer should return 1 unit and not 2. It didn't run in my colab at all, until I changed the last dense later to 1 from 2.
I was getting some error from keras because of that.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Hi Sir, its not necessary that all the input images will be of size 224 * 224, some of the images which I have downloaded from google are far less than the said size, how to handle such images. Please guide.
YOU can reshape the size of all the images in a folder, with the various preprocessing functions, and all the images in a folder will be resized to desired size and then deed it.
Sir, can prepare video lecture on image segmentation using deeplearning in keras like UNet
Sure and soon.
fit_generators method is not working for sequential model
How i use vgg16 If the shape of image is not square like 640x480 ?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
sir how to calculate precision and recall and confusion matrix
Great work sir
Thanks Manju
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Sir, why channel size is increasing?
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
check out the details
Sir actually,ResNet,VGG16,AlexNet implement by which software (i.e)matlab or python....anybody please clarrify
Python with many libraries being used.
Is That will not work on matlab??
@@zenonisha505 it will.. you need to pull right module
python . that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Can You Provide the PPT used in this video
Let me check
@@ShriramVasudevan thank you very much
Sir I am getting an error saying shapes incompatible(25088,256) and (25088,4096) So can u share the correct vggweights.h5 file
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Ma lord and savior has bless me
helo, good video. But can you teach how to implement vgg16 with 6 different classes? Where we should edit in the code? Thank you for your time
I shall make a quick video on that
Thank you so much Sir.. I appreciate that.
Hope you can release the video quickly since I am using your explanation as one of my references in my final year project 😁
Sir I am getting error as module leraa has no attribute applications
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
Can I get your mail as I have few questions regarding this topic??
Very nice 👍
Thank you
brother ppt yet not given
I get this error : OSError: Unable to open file (unable to open file: name = 'vggweights.h5', errno = 2, error message = 'No such file or directory', flags = 0, o_flags = 0)
How do I find out the path location to the downloaded VGG code?
You need to download the Weights file. . And keep it on the path. It woll work fine thenn
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
@@ShriramVasudevan how to downlode weights file?
Can u plzz explain GMCNN with code...
Please ...
Thank you
Soon
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
can u Share the code please
Can u share code with us plz
By the way awesome explanation
It's in my git. Shriramkv
Thanks and glad u liked
@@ShriramVasudevan sir one more question at dense layer units is 4096 , so why we use only 256 units
@@sagarprithvipalsinghrawat6655 We are only using top 13 layers for our model from VGG16. Classification part(After flattening) is upto us. We can use any number of neurons
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19
good video, but this the first time that the indian accent was so strong i had difficulties understanding parts
pls share the github link for the code
@shriramkv
sir can u copy paste the code here
its in my git. @shriramkv
can you pl share this script ....or please share github link if you have
It's in GIT
Shriramkv
could you please add subtitle?
Will that as soon as I can
@@ShriramVasudevangreat video, helped me a lot from a different background.
that is why I said this video provides wrong information this video doesn't even provide the full form of Vgg please ua-cam.com/video/PGhbpnfC3Qc/v-deo.html it covers bgg 16 and says that it is bgg 19