This repository try to model the neural response in visual cortex by features encoded in well trained CNNs.
Features the following methods
- Factorized regression model
- Correlation feature tensor model
- Partial least square model
It provides utilities around this purpose.
- Loading experimental data, pairs of neural response and image or path to image.
In additions it contains analysis code for defining pixel wise objectness of an image.
- Summary the correlation or weight tensor along the spatial dimensions or channel dimensions.