TensorFlow Keras Model with Mixed Inputs Tutorial
HTML-код
- Опубликовано: 20 авг 2024
- The Colab Notebook: colab.research...
Thank you for watching the video! You can learn data science FASTER at mlnow.ai!
Master Python at mlnow.ai/cours...!
Learn SQL & Relational Databases at mlnow.ai/cours...!
Learn NumPy, Pandas, and Python for Data Science at mlnow.ai/cours...!
Become a Machine Learning Expert at mlnow.ai/cours...!
Don't forget to subscribe if you enjoyed the video :D
Take my courses at mlnow.ai/!
Thank you very much ! As a college student who just started getting a grasp of the code behind these models this project was very insightful.
great work and explanation much needed one
this is very helpful and insightful, thank you very much (you've earned a new sub btw)
Glad to hear it!
Very helpful thanks!!!
You're very welcome Arsheya 😃😃
great insight! multimodal usecase solved
hey man thank you very much! great tutorial :) subbed!
You're very welcome, and thank you :)
I rally love your channel and your explanation, it's like to find the answer whenever I find a video for you
I have a question
input = whatever
x = whatever (input)
y = whatever (x)
is this correct for the parentheses goal ? it represent the previous layer
Not so traditaional, I like that.
Can we talk about a little doubt? Im tryng do to similar process but having some problems
can anyone tell me how to get the files he uses in the video?
He even looks like a Pokemon trainer
Nice Explanation; I found an error like this "Layer "model_2" expects 2 input(s), but it received 1 input tensors. Inputs received: []" while fitting the model. Can you figure out the problem here?
Hi, Nice work. Appreciate your time. I have one problem when I scale it. I have 200000 images, each size is 224,224,3 , and after loading that when I use np.array the colab is crashed due to lack of memory. So, how to deal with such a situation?
Ronak you can use Batchloader, stochastic gradient descent or downsample images even further or in colab set GPU target.
Hi Greg great tutorial! Could you possibly guide me on understanding how i could use a model created like this, to have just an image as an input to output a predicted 'HP'. Thanks.
Edit: I misunderstood and realised that this model cannot take an input of just an image to output the predicted 'HP', however what is the process in order for it to do so, just removing the feed-forward stream would suffice?
Thank you! It's a lot easier to have just an image as the input, for this you can see any such tutorial on CNN's :)
Great Video! I am trying to learn ML in order to create a similar model. I was wondering if you what kind of role did the picture play? My model should consider the colors present in the picture and I didnt exactly got what was the role of the picture in this case
To be honest the picture probably didn't help here, mostly just a fun example to show that this is possible. There's definitely times, especially with medical data, these these types of data could be paired together. And thanks for the kindness!!
Hi Greg ! do you think is possible to use something like vectorization instead of iterrows() for the creation of the npz ?
Yeah, I'm sure you can - great find!
Thank you for this. I've been stuck for a while and this one somehow works. I have a question though (maybe the answer is already in the video and I've missed it), does storing the image in npz format and basic jpg/png format have any difference when used in training the model like longer/shorter training time, lesser RAM usage, or something?
You're welcome! They way I setup the model part, no, it doesn't use the npzs to use less ram. It still loads everything in memory. That is indeed why I showed the npzs though, because there is a way to make TensorFlow only pull in the files that it happens to need for each batch, to save Ram.
For more info, lookup TensorFlow datasets :)
Hi, thanks for this tutorial. Is mixed input also applicable to classification problem?
Absolutely.
@@GregHogg Thanks for your reply. Is this concept similar to something called multi-head / multi-input CNN ? The only difference here is that all the input of multi-input CNN are images, but the inputs of mixed-input model contains both image and structural dataset (csv).
Hi Greg, interesting video but how can I use flow_from_directory in Keras with muti inputs model like the one you showed. Thanks.
I don't know if you can mix generators with normal data - but I would try to do what I did and run through the data first to be safe.
@@GregHogg Can you please advise how I can use that function to connect it with that neural network model with 2 inputs images and numerical data. Thanks.
@@chandasimfukwe4555 I just replied and said I'm not sure if it will work and suggested to do it closer to how I did.
@@GregHogg Thanks Greg, it has really helped me. Looking forward to your future videos.
@@chandasimfukwe4555 Super glad to hear that.