https://github.com/i008/deepcounting

neural networks are cool but can they count?
https://github.com/i008/deepcounting

convolutional-neural-networks data-augumentation data-generation

Last synced: about 2 months ago
JSON representation

neural networks are cool but can they count?

• Host: GitHub
• URL: https://github.com/i008/deepcounting
• Owner: i008
• Created: 2017-02-15T17:57:51.000Z (over 9 years ago)
• Default Branch: master
• Last Pushed: 2017-03-09T21:12:29.000Z (over 9 years ago)
• Last Synced: 2025-08-14T19:28:36.956Z (11 months ago)
• Topics: convolutional-neural-networks, data-augumentation, data-generation
• Language: Jupyter Notebook
• Size: 3.99 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.MD

Awesome Lists containing this project

README

          
### Idea

The question is simple, what happens if we generate images with some kind of objects on them (circles for example) and train 

a CNN giving it only the count of the objects as the target variable. Will it learn to count? will it learn to segment properly?  

The biggest challange of the image-generation part is to avoid data-leakage  for example the network fitting to the total-area of the foreground instead of the object count.