Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/tcapelle/torch_moving_mnist
A simple Dataset generator for Moving Mnist
https://github.com/tcapelle/torch_moving_mnist
Last synced: 29 days ago
JSON representation
A simple Dataset generator for Moving Mnist
- Host: GitHub
- URL: https://github.com/tcapelle/torch_moving_mnist
- Owner: tcapelle
- License: apache-2.0
- Created: 2023-01-10T20:53:19.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2023-05-26T09:10:25.000Z (over 1 year ago)
- Last Synced: 2024-08-01T16:45:59.901Z (3 months ago)
- Language: Jupyter Notebook
- Homepage: https://tcapelle.github.io/torch_moving_mnist
- Size: 6.83 MB
- Stars: 9
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
torch_moving_mnist
================
## Install
``` sh
pip install -e .
```
## How to use
``` python
from types import SimpleNamespace
from torch_moving_mnist.data import MovingMNIST
from torch_moving_mnist.utils import show_images
```
``` python
affine_params = SimpleNamespace(
angle=(-5, 5), # rotation in degrees (min and max values)
translate=((-5, 5), (-5, 5)), # translation in pixels x and y
scale=(.9, 1.1), # scaling in percentage (1.0 = no scaling)
shear=(-2, 2), # deformation on the z-plane
)
```
Create a MovingMNIST dataset with `affine_params`, with 10 frames and
may include up to 3 digitis. Image size is 64.
``` python
from nbdev.showdoc import show_doc
```
------------------------------------------------------------------------
### MovingMNIST
> MovingMNIST (path='.', affine_params:dict=namespace(angle=(-4, 4),
> translate=((-5, 5), (-5, 5)), scale=(0.8, 1.2), shear=(-3,
> 3)), num_digits:list[int]=[1, 2], num_frames:int=4,
> img_size=64, concat=True, normalize=False)
Initialize self. See help(type(self)) for accurate signature.
| | **Type** | **Default** | **Details** |
|---------------|----------|-----------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------|
| path | str | . | path to store the MNIST dataset |
| affine_params | dict | namespace(angle=(-4, 4), translate=((-5, 5), (-5, 5)), scale=(0.8, 1.2), shear=(-3, 3)) | affine transform parameters, refer to torchvision.transforms.functional.affine |
| num_digits | list | \[1, 2\] | how many digits to move, random choice between the value provided |
| num_frames | int | 4 | how many frames to create |
| img_size | int | 64 | the canvas size, the actual digits are always 28x28 |
| concat | bool | True | if we concat the final results (frames, 1, 28, 28) or a list of frames. |
| normalize | bool | False | scale images in \[0,1\] and normalize them with MNIST stats. Applied at batch level. Have to take care of the canvas size that messes up the stats! |
``` python
ds = MovingMNIST(affine_params=affine_params, num_frames=10, num_digits=[1,2,3], img_size=64)
```
when you index the dataset, it generates a random set of MNIST digits
and trajectories. You could basically only call `ds[0]`
``` python
sequence = ds[0]
```
``` python
show_images(sequence, figsize=(20,10))
```
``` python
t = sequence
type(t), t.shape
```
(torch.Tensor, torch.Size([10, 1, 64, 64]))
## Dataloader
This dataset is randomly creating sequences on the fly, so the
dataloader is just going to generate a batch…
``` python
batch = ds.get_batch(bs=128)
```
``` python
batch.shape
```
torch.Size([128, 10, 1, 64, 64])
``` python
show_images(batch[0])
```
the dataloader also is normalizing the inputs for you, after
constructing the batch.
``` python
ds.batch_tfms
```
Compose(
ConvertImageDtype()
)