| #!/bin/bash | |
| # Paste at Pastebin.com using command line (browsers are slow, right?) | |
| # coder : Anil Dewani | |
| # date : November 7, 2010 | |
| # forker : Igor Duarte Cardoso | |
| # date : March 12, 2014 | |
| # Don't forget to set your Unique Developer API Key here | |
| # http://pastebin.com/api#1 | |
| API_KEY= |
This is a tribute to Studio Ghibli's Exhibition in Paris.
While walking through the exhibit, I learned that the castle in howl's moving castle was actually animated with a computer. I was surprised, and inspired to recreate this famous scene using web technologies that I'm familiar with, allowing the user to control the castle's movement.
I ended up biting off more than I could chew, and spending much longer than I had initially intended.
To achieve it, I cut all of the elements out of the film itself using photoshop to crop certain frames. I then inserted all of the elements directly to the DOM (no canvas) and animated them using GSAP and an epic amount of timelines. The legs' walk cycle were by far the most difficult part. I did not use any other software for the animation, just good old sublime text!
Just you, your code and nothing else.... :meditatingbuddha:
| """ | |
| Create train, valid, test iterators for CIFAR-10 [1]. | |
| Easily extended to MNIST, CIFAR-100 and Imagenet. | |
| [1]: https://discuss.pytorch.org/t/feedback-on-pytorch-for-kaggle-competitions/2252/4 | |
| """ | |
| import torch | |
| import numpy as np |
| import torch | |
| from torchvision import datasets | |
| class ImageFolderWithPaths(datasets.ImageFolder): | |
| """Custom dataset that includes image file paths. Extends | |
| torchvision.datasets.ImageFolder | |
| """ | |
| # override the __getitem__ method. this is the method that dataloader calls | |
| def __getitem__(self, index): |
