What is Transfer Learning? | With code in Keras

Рет қаралды 23,665

Күн бұрын

What is transfer learning? Why do people use it to develop projects? Let's learn all about it in this video.
In the second part, we also see how to implement an example transfer learning model using the Xception model in the Keras Applications library with a dataset from TensorFlow datasets.
Find the notebook here to follow along with the tutorial: colab.research.google.com/dri...
Get your free AssemblyAI API token here 👇
www.assemblyai.com/?...
▬▬▬▬▬▬▬▬▬▬▬▬ CONNECT ▬▬▬▬▬▬▬▬▬▬▬▬
🖥️ Website: www.assemblyai.com
🐦 Twitter: / assemblyai
🦾 Discord: / discord
▶️ Subscribe: kzfaq.info?...
🔥 We're hiring! Check our open roles: www.assemblyai.com/careers
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
#MachineLearning #DeepLearning

Пікірлер: 34

@TechRevelation2024 13 күн бұрын

Your videos are so informative and knowledgeable . Can you make a video on hugging face models with API connections

@chrstfer2452 7 ай бұрын

It'd be sick if you put links in the description of your shorts too, but glad to find them here. Mad nice, btw, subbed.

@bhargavamahesh 8 ай бұрын

Very nicely explain the Transfer Learning. I want more videos in this line.....so keeping posting latest TL video

@patloeber 2 жыл бұрын

Awesome tutorial!

@AssemblyAI 2 жыл бұрын

Thanks Python Engineer!

@nogur9 11 ай бұрын

Thanks :) it's great

@user-ce8jy4jc4f 9 ай бұрын

we would like more transfer learning and fine tuning vids thanks

@hudaali6596 Жыл бұрын

nice information 🌷

@AssemblyAI Жыл бұрын

Thanks!

@md.alamintalukder3261 Жыл бұрын

Love it

@AssemblyAI Жыл бұрын

Great!

@nikolamatijasevic9454 Жыл бұрын

I think this is the best tutorial on KZfaq! Just one question, can I load a custom dataset to use or it has to be found in the tfds? Thanks!

@AssemblyAI Жыл бұрын

Thank you! You can use a custom dataset too of course. Just need to make sure it is the correct format . :)

@keithwins 6 ай бұрын

Thank you

@harshays2873 Жыл бұрын

Need some Explanation about Optimizers also.

@tilkesh Жыл бұрын

Thx

@user-rn5jl6bz8o Жыл бұрын

@AssemblyAI Good tutorial, I'll like to use this for my problem dataset to improve accuracy/detection. My challenge is I have about 60 Png images that are not particularly distinguishable, wondering if this method will work well for me. Any suggestions?

@ASdASd-kr1ft Жыл бұрын

Great tutorial, can explain the same but for pytorch? Thanks!. Sorry by my English

@CheapSheap Жыл бұрын

She explains beautifully, what is her name?

@shivayshakti6575 Жыл бұрын

Sandy

@mitolojikitabi 3 ай бұрын

Sen nasıl bir yavrusun ya, ara ara karşıma çıkıyon bakmak zorunda kalıyorum.

@leonardovoltolini2990 Жыл бұрын

I always see transfer learning applied to image recognition, but I would like an example of it being applied to predicting "numbers". Does anyone have an example?

@user-tg4te4co3y Жыл бұрын

@AssemblyAI yeah, I too have been looking for the application of TL in a regression use case. Do you have any tutorials for this?

@lotfiamr8433 Жыл бұрын

@@user-tg4te4co3y mee too did you find anything please ?

@lotfiamr8433 Жыл бұрын

@@user-tg4te4co3y ?

@bigdoor64 2 жыл бұрын

Hi! At 8:23, you say "to catchup with the model". Could you explain what this means in further detail please? I've not heard that term used before!

@AssemblyAI 2 жыл бұрын

It’s not really a technical term, I used it to give an intuitive understanding. What I meant with it is that when you’re working with a pre-trained model, it is already good at a certain task. With transfer learning you add an output layer that is randomly initialized so it is not yet good at producing good results. By running the pre-trained model and the freshly created output later together you let the output layer train to a level that performs to an acceptable level. During that time, the pre-trained model’s weight are not updated. And once the model performs well enough (this would depend on your model’s goal performance) it means the new output layer has caught up. Hope this clears it up! - Mısra

@bigdoor64 2 жыл бұрын

@@AssemblyAI Thank you!

@TuhinBhattacharya 2 жыл бұрын

Awsome.. Can we have similar in pytorch?

@AssemblyAI 2 жыл бұрын

We'll add that to our video planning. Thanks for the input! - Mısra

@OccamsPlasmaGun 2 жыл бұрын

how would one do transfer learning with a keypoint detector? The few examples out there are obsolete and with conflicting dependencies.

@AssemblyAI 2 жыл бұрын

Hey Andrew, I don't have experience with keypoint detection. Have you searched the pre-trained model libraries to see if they have a useful pre-trained model? -Mısra

@OccamsPlasmaGun 2 жыл бұрын

@@AssemblyAI of course. There are none that are easily adaptable. I'm using one from github, but can't get the environment set up so that it works. Python sucks even more than node.js