https://github.com/javierantoran/corner_dataset

Simple dataset for viewing internal neural net representations
https://github.com/javierantoran/corner_dataset

Last synced: 6 months ago
JSON representation

Simple dataset for viewing internal neural net representations

• Host: GitHub
• URL: https://github.com/javierantoran/corner_dataset
• Owner: JavierAntoran
• Created: 2018-03-23T19:19:38.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2018-10-07T14:40:44.000Z (almost 7 years ago)
• Last Synced: 2025-02-14T20:37:55.383Z (8 months ago)
• Language: Python
• Homepage:
• Size: 30.1 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# corner_dataset

Simple dataset used for viewing how a multi layer perceptron organises

information throughout its activation space.

The dataset consists of 4096 32x32 pixel images belonging

to 4 different classes. Each class shows two lines forming a right angle

pointing towards a different corner of the image. Line length, thickness and possition vary across images. Low amplitude

gaussian noise is added to images.

The dataset can be loaded using numpy:

```python

x = np.load('data/angle_ims.npy')

y = np.load('data/angle_targets.npy')

```

 ![Alt text](pics/100_samples.png?raw=true "some examples")