I can't enroll in deep learning paid courses due to my weak financial conditions but due to Xcampus I am able to learn deep learning :') This channel is a blessing for peoples like me
since 2 month I was trying to find out the difference between finetuning and feature extraction but there were no such video that can explain this 2 topic in this easy manner, thanks to you Nitish
Mark my word! This channel will be the best channel for data science community. No PR just pure content. People will flow lessely follow this channel just need to here once. Keep up the good work man.
I already learned these concept but jo clarification aap ki video se aati he wo ab kisi english video se nhi. The Best Machine Learning channel in Hindi/Urdu. Love from Pakistan.
00:02 Transfer learning is crucial due to data and time constraints in training deep learning models. 03:14 Transfer learning allows for using pre-trained models in your projects. 06:40 Transfer learning enables leveraging pre-trained models for new tasks 09:21 Transfer learning applies knowledge from one domain to another 12:26 Transfer learning involves using pre-trained model knowledge in a new task 15:49 Transfer learning involves extracting features and fine-tuning for specific tasks. 19:03 Transfer learning involves fine tuning the last layers. 22:28 Implementing Transfer Learning in Keras 26:03 Reducing overfitting through data augmentation 28:48 Transfer learning in Keras involves fine-tuning and feature extraction. 31:15 Transfer learning involves fine tuning and feature extraction 33:40 Transfer learning can greatly enhance learning Crafted by Merlin AI.
Thanks Nitish. this video is superb. if anyone is getting the error , import these libraries. (keras library is changed for load_img and img_to_array from keras.utils import img_to_array, load_img from keras.applications.vgg16 import preprocess_input from numpy import expand_dims also use "keras.Model" instead of only "Model" to get the layers, use this --> model = keras.Model (inputs=model.inputs,outputs=model.layers[1].output) instead of -->model = Model (inputs=model.inputs,outputs=model.layers[1].output)
I have a question. When u applied the data augmentation why did not no of images in train and test folder increase.? the no of files were same before and after the data augmentation part.
absolutely brilliant even though I have the detailed concept of the CNN all families alogs but still this video help to understand the coding structure of it it is really ammazing
Hi, I like ur videos so much as these are very effective to learn the foundation of any topic.Tnx for the great efforts. I have a question. If image net data has already dog and cat class, why you retrained again? If I want to find similar objects like mobile which looks similar based on a object input like any mobile but these type object is not included in image net data set I mean mobile then what is the best way to use teansfer learning to get the job done?
Tried the fine tuning code for vgg16 pretrained model on multiclass dataset. #no. of wrong predictions wrong =0 for i in range(len(y_true)): if(y_true[i] != y_pred[i]): wrong = wrong+1 print("total wrong predictions: ",wrong) print("total correct predictions: ", (len(y_true)-wrong)) Gave output as: total wrong predictions: 1843 total correct predictions: 38 Although the validation accuracy is nearly 96%, but still the predictions are wrong. Any reason why???
What is difference between Continual Learning and Transfer Learning? Also, What if I Want to implement continual or Transfer learning but I don't have pre-trained model like u mentioned VGG16,Restnet (So, Dataset might me in this case would be in tabular form e.g CSV data) Do you have any notes or code for continual learning on Tabular data?
I'm following the steps as explained and training on my dataset but in all the epochs the training and validation accuracy is coming constant . My dataset is of images and has 3 classes with total of 2000 training and 200 validation images. Can someone please help in identifying where the problem is.
[Very Urgent] Please tell what can we call these implementations, if applied to a disease(say covid) dataset problem: Covid detection, covid classification, covid diagnosis, covid prediction ya automated diagnosis/ automated detection. What exactly do we call it when we use/implement this code. i have no idea. Please kindly tell. People from comments also can advice. Please it is very urgent 🙏🙏
your video are very helpful for machine and deep learning. kindly guide me can we use this model for medical images like lungs cancer etc. and plz guide for multi class classification. if you have any video on multiclass classification with the help of transfer learning, kindly share its link
I graduated in the year of 2020. Completed data science course and learning more concepts through this channel, I am still fresher, does my graduated year affect for getting job in this field?
Sir I worked on kidney stones dataset, and same I faced overfitting problem . Train accuracy in 80s and test accuracy in 40s . But I was unable to overcome.
Sir, Can I do transfer learning on my own custom "model" that is created by me? That is I have created and trained my own CNN model saved as xyz.keras. I now want to do transfer learning of this model on new type of data. Is that possible?
Thanks a lot for your all efforts and hardwork you do for us🙏🙏 Thanks for python playlist .. Sir , can you tell me that what i should do after completing your python playlist ( in details) ?? Please....complete your 100 days of python playlist or you will tell us that what topics are covered in 100 days of python playlist so we can prepare.. How can I be machine Learning Engineer or data Scientist in FACEBOOK/ GOOGLE/ AMAZON/ APPLE/ MICROSOFT ?? Sir....please....please...reply me 🙏🙏🙏🙏🤟🥳
Our teacher told us that transfer learning is used in CNNs because initially features are somewhat similar and change in the later layers. So, my question is whether transfer learning can only be used in CNN tasks? Please Reply
What a great explanation. I have a question. How did you train 25000 images in Google Collab? Because When I use free Google Collab on 10000 images, all GPU units are utilized. What is the solution?
I ran the last notebook but results are not good....accuracy is 50%. I didn't make even a single change. Please have a look at the last one in which you did finetunning again. Thanks
I can't enroll in deep learning paid courses due to my weak financial conditions but due to Xcampus I am able to learn deep learning :') This channel is a blessing for peoples like me
Same brother
Deep learning paid course 😮
Same
Its campusX
guys apply for financial aid in coursera
since 2 month I was trying to find out the difference between finetuning and feature extraction but there were no such video that can explain this 2 topic in this easy manner, thanks to you Nitish
आपकी तारीफ मे
मै इतना समझदार और होशियार हू तोह मे क्या करू youtube channel खोल के मुफ्त मे ज्ञान फैलाउ
Thanks brother
this channel is just gold for machine learning enthusiasts
Mark my word! This channel will be the best channel for data science community. No PR just pure content. People will flow lessely follow this channel just need to here once. Keep up the good work man.
I already learned these concept but jo clarification aap ki video se aati he wo ab kisi english video se nhi. The Best Machine Learning channel in Hindi/Urdu. Love from Pakistan.
I think campus x needs more subscribes and his explanation was great.no other channels can not explain in this way....
This lecture deserves 1 million views.
I just love the way you simplified the concepts like transfer learning, fine tuning, and the architecture of the CNNs. Thank you bhaiya🙏
The most amazing thing is I come to watch your videos to get the cream layer of the knowledge
just amazing bro... ❤️.. even 10th class student can understand your explanation... straight simple clear
I am fortunate to have found your channel. Phenomenal work that you have been doing in democratising ML knowledge. Thank you Nitish!
No other channels explained this topic better than your explanation.
Your teaching style and concepts are tremendous. Thank you sir.
Very Good One of the best videos on deep learning
best video on UA-cam channel thanks for sharing
The hard work you put into your videos is amazing.
00:02 Transfer learning is crucial due to data and time constraints in training deep learning models.
03:14 Transfer learning allows for using pre-trained models in your projects.
06:40 Transfer learning enables leveraging pre-trained models for new tasks
09:21 Transfer learning applies knowledge from one domain to another
12:26 Transfer learning involves using pre-trained model knowledge in a new task
15:49 Transfer learning involves extracting features and fine-tuning for specific tasks.
19:03 Transfer learning involves fine tuning the last layers.
22:28 Implementing Transfer Learning in Keras
26:03 Reducing overfitting through data augmentation
28:48 Transfer learning in Keras involves fine-tuning and feature extraction.
31:15 Transfer learning involves fine tuning and feature extraction
33:40 Transfer learning can greatly enhance learning
Crafted by Merlin AI.
Thank you Nitish for "Transferring your Learning" to all the students and learners who wish to be a deep learning expert....
SIR HEAD OFF TO YOU THIS WAY YOU UNDERSTAND CONCEPT SO EASILY DIE HEART FAN OF YOU
Bhai aaap ko bohot saari doooaaye, kya tha main , kya ban gya aapse sikh ke. god bless you :)
the way u teach is just fab sir
one of the best videos of the list...Campusx u rock
Due to your best way of teaching, I subscribed your channel and I hope you will upload more useful and knowledgeable videos.
Thankyou very much sir,I don't have word to say thank you,unbelivable lecture sir.❤❤❤❤❤
I pressed the like button before watching the video because I knew, it would be the best one.
no matter how much thankyou i say it will still be less. really sir thankyou so much.
You are great teacher sir.❤❤
College Wale to free ki fess le rahe hai
Really Loved the entire explanation about Transfer Learning!!🔥
Amazing explanation sir
thanks a lot for provinding free content
Thanks Nitish. this video is superb.
if anyone is getting the error , import these libraries. (keras library is changed for load_img and img_to_array
from keras.utils import img_to_array, load_img
from keras.applications.vgg16 import preprocess_input
from numpy import expand_dims
also use "keras.Model" instead of only "Model" to get the layers,
use this --> model = keras.Model (inputs=model.inputs,outputs=model.layers[1].output)
instead of -->model = Model (inputs=model.inputs,outputs=model.layers[1].output)
Nice video! Btw your mic is so good I got bluffed twice by the background noise.
Maza aa gya sir...!! Transfer Learning is Over Powered OP🔥🔥🔥🔥🔥🔥!!!
Really very nice video (Both explanation and coding part ) Sir!
Thank you very much!
wow, explained all the things in very easy manner !!
Thankyou very much sir, amazing and total understandable lecture sir. 🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏
Great no words for such extraordinary skill
thank you so much sir, you have cleared my all doubts related to transfer learning..
I have a question. When u applied the data augmentation why did not no of images in train and test folder increase.? the no of files were same before and after the data augmentation part.
absolutely brilliant even though I have the detailed concept of the CNN all families alogs but still this video help to understand the coding structure of it it is really ammazing
sir aap ko dekh k feeling blessed 🙂
Sir please make video on how to become a data scientist or complete roadmap to become a data scientist
thank you for too much information which I wanted to gain ❤❤
best tutorial ever on transfer learning....😍😍
You explained so well..👍
ThankYou so much sir for amazing Explanation.
Love from Bangladesh ❤️
Great and honest teaching
This is really awesome .. Aisey hi hum kya deeplabv3 aur mobilenetv3 bhi pretrained models use kar saktey hain kyaa..?
Hi NItish, Can you also cover RCNN, Fast-RCNN, Faster-RCNN & YOLO Algorithm in CNN.
It would be of great help. Thanks!!
Sir great explanation!
Very good explanation sir.
Hi, I like ur videos so much as these are very effective to learn the foundation of any topic.Tnx for the great efforts.
I have a question.
If image net data has already dog and cat class, why you retrained again? If I want to find similar objects like mobile which looks similar based on a object input like any mobile but these type object is not included in image net data set I mean mobile then what is the best way to use teansfer learning to get the job done?
Please start mentorship program.. And also guide me for PhD in NLP domain..
I am interested too
In the 3rd notebook im getting some problem
The accuracy is just stuck around 50% for some reason
@campusX where are the label file in the cats vs dog dataset I can see only Image files in it.
WHAT ARE THE PLANS FOR RESTARTING THE MENTORSHIP PROGRAM?
I am interested too
Sir plz make a video on LSTM, your video helps me a lot .... thank you
how do i use this for multiclass classification , tried it but got lower accuracy
Nicely Explained , Thanks
Tried the fine tuning code for vgg16 pretrained model on multiclass dataset.
#no. of wrong predictions
wrong =0
for i in range(len(y_true)):
if(y_true[i] != y_pred[i]):
wrong = wrong+1
print("total wrong predictions: ",wrong)
print("total correct predictions: ", (len(y_true)-wrong))
Gave output as: total wrong predictions: 1843
total correct predictions: 38
Although the validation accuracy is nearly 96%, but still the predictions are wrong.
Any reason why???
What is the difference between Transfer learning and domain adaptation?
What is difference between Continual Learning and Transfer Learning?
Also, What if I Want to implement continual or Transfer learning but I don't have pre-trained model like u mentioned VGG16,Restnet (So, Dataset might me in this case would be in tabular form e.g CSV data)
Do you have any notes or code for continual learning on Tabular data?
How many of the layers we should unfreeze in fine tuning?
I have one question ❓
Which books you are preferred to get this deep knowledge of each and every topic.
Please reply me sir with your valuable answer.
Please make vidoes on transfer learning methods for NLP
Can you apply this on UCIHAR dataset? That's a sensor data which u have worked also in feature selection
I'm following the steps as explained and training on my dataset but in all the epochs the training and validation accuracy is coming constant . My dataset is of images and has 3 classes with total of 2000 training and 200 validation images. Can someone please help in identifying where the problem is.
bhai maja agaye ,kya smajhaya h apne
[Very Urgent] Please tell what can we call these implementations, if applied to a disease(say covid) dataset problem: Covid detection, covid classification, covid diagnosis, covid prediction ya automated diagnosis/ automated detection. What exactly do we call it when we use/implement this code. i have no idea. Please kindly tell. People from comments also can advice. Please it is very urgent 🙏🙏
Hello Sir, Please also make videos on BERT model.
your video are very helpful for machine and deep learning. kindly guide me can we use this model for medical images like lungs cancer etc. and plz guide for multi class classification. if you have any video on multiclass classification with the help of transfer learning, kindly share its link
Sir we can use unsupervised learning for label the Dataset and then perform cnn? Help me please sir
sir ham fine tuning ke sath data augmentation use karsakte ha
I graduated in the year of 2020. Completed data science course and learning more concepts through this channel, I am still fresher, does my graduated year affect for getting job in this field?
completed ,. Awesome content
i reached 98.98% accuracy as well as val_accuracy 98.16% with Inception V3
can we apply transfer learning on tabular data?
Sir I worked on kidney stones dataset, and same I faced overfitting problem . Train accuracy in 80s and test accuracy in 40s .
But I was unable to overcome.
I LOVE THIS GUY
Hey, can you please share the link to the resources that you use while recording? For instance the scribble pad and the software.
how to use alexnet as pretrained model
Hello, I found this content useful if I have to use xception incase of VGG-16 then how to use, which layers to freeze.
I'm Doing this same thing with Pytorch plz any help me how do i fine tune
can we apply the same code for the transfer learning of text sentiments ?
Sir, Can I do transfer learning on my own custom "model" that is created by me? That is I have created and trained my own CNN model saved as xyz.keras. I now want to do transfer learning of this model on new type of data. Is that possible?
Thanks a lot for your all efforts and hardwork you do for us🙏🙏
Thanks for python playlist ..
Sir , can you tell me that what i should do after completing your python playlist ( in details) ??
Please....complete your 100 days of python playlist or you will tell us that what topics are covered in 100 days of python playlist so we can prepare..
How can I be machine Learning Engineer or data Scientist in FACEBOOK/ GOOGLE/ AMAZON/ APPLE/ MICROSOFT ??
Sir....please....please...reply me
🙏🙏🙏🙏🤟🥳
Hello sir, If has been pretrained on a text classification model, can we use this model in text generation, question answering or paragraph summary?
Good explanation
Unmatchable 😊
Our teacher told us that transfer learning is used in CNNs because initially features are somewhat similar and change in the later layers. So, my question is whether transfer learning can only be used in CNN tasks?
Please Reply
Drop a mail to them
NLP playlist pending
What a great explanation. I have a question. How did you train 25000 images in Google Collab? Because When I use free Google Collab on 10000 images, all GPU units are utilized. What is the solution?
Bhai Seq2Seq ki video kab aye ge ?
All I need .....just found in one video
I ran the last notebook but results are not good....accuracy is 50%. I didn't make even a single change. Please have a look at the last one in which you did finetunning again. Thanks
I guess it is giving good results because it has been trained on cat dog images
Great video❤
Sir plz continue machine learning playlist
Great explanation