JGuymont / humanware

Different convolutional neural network implementations for predicting the lenght of the house numbers in the SVHN image dataset. First part of the Humanware project in ift6759-avanced projects in ML.

Geek Repo

Github PK Tool

deep-learning convolutional-neural-networks residual-networks squeeze-and-excitation pytorch university-project svhn-classifier svhn-dataset

link to the overleaf report

How to use

Execute python split_data.py to split the training set metadata in training/validation/test (default 0.7/0.2/0.1)
Put your model's class in the folder models.
Put all the settings you want in a ini file in config/ (check *config/example.ini for an example of configuration).
Execute python main.py config/<name>.ini to start the training of your model.

About

Different convolutional neural network implementations for predicting the lenght of the house numbers in the SVHN image dataset. First part of the Humanware project in ift6759-avanced projects in ML.

deep-learning convolutional-neural-networks residual-networks squeeze-and-excitation pytorch university-project svhn-classifier svhn-dataset

Languages

Language:Python 95.7%Language:Shell 4.3%