L17.6 A Variational Autoencoder for Face Images in PyTorch -- Code Example

Ғылым және технология

Slides: sebastianraschka.com/pdf/lect...
L17 code: github.com/rasbt/stat453-deep...
Discussing 2_VAE_celeba-sigmoid_mse.ipynb,
3_VAE_nearest-neighbor-upsampling.ipynb
& 4_VAE_celeba-inspect-latent.ipynb
-------
This video is part of my Introduction of Deep Learning course.
Next video: • L17.7 VAE Latent Space...
The complete playlist: • Intro to Deep Learning...
A handy overview page with links to the materials: sebastianraschka.com/blog/202...
-------
If you want to be notified about future videos, please consider subscribing to my channel: / sebastianraschka

Пікірлер: 9

  • @geetagoel1540
    @geetagoel15403 жыл бұрын

    Hi Sebastian, Great explanation! This is really amazing how you are making everything looks so easy. Your course is doing wonders to the beginners. Thank you so much to make it public and let us learn painlessly.

  • @SebastianRaschka

    @SebastianRaschka

    3 жыл бұрын

    Thanks so much for the kind works. Happy to hear that it's useful :)

  • @goopakaua5999
    @goopakaua59992 жыл бұрын

    Hello, can you please inform how to get the helper_train (and train_vae_v1) ? On trying to pip install it says no matching distribution found..have been searching on net, no luck yet!

  • @SebastianRaschka

    @SebastianRaschka

    2 жыл бұрын

    Oh sorry about the confusion, you don't have to install anything. You just need to have the helper_train file in the same folder as the code notebook or script. E.g., have a look at the link in the KZread description (github.com/rasbt/stat453-deep-learning-ss21/tree/main/L17). If you organize your code like this, it should work. Or, if you want to put the helper file somewhere else, let's say you save it under "/Users/yourname/Desktop/helper_train.py" what you can do is adding the following lines on top of the notebook: import sys sys.path.insert(0, "/Users/yourname/Desktop/")

  • @thienan7206
    @thienan72064 ай бұрын

    Hi, I know it has been 2 years from the release of the video. But can I ask, here your loss calculated is quite small, however, when applying for a big set of images in shape 160 x 320 x 1 (grayscale images only), the loss is always run out of control, and turn to Nan type in the early stage of training. Do you have any thoughts on this ? It would be great to receive your answer

  • @mistervoldemort7540
    @mistervoldemort7540 Жыл бұрын

    Hi, thanks for sharing, great video. What LEARNING_RATE value did you use ?

  • @SaschaRobitzki

    @SaschaRobitzki

    5 ай бұрын

    Should be `LEARNING_RATE = 0.0005`.

  • @mistervoldemort7540

    @mistervoldemort7540

    5 ай бұрын

    @@SaschaRobitzki thanks

Келесі