Classifying different scenarios (city, desert, mountain, nature, sea, universe).
Project built using Python: Keras (with Tensorflow backend), numpy and scikit-learn.
Dataset structure
The dataset is made of six folders (city, desert, mountain, nature, sea, universe), each one containing 25 pictures. An example of city picture is provided below.
How to install it
$ git clone https://github.com/marcogdepinto/ScenarioClassifier.git
$ pip install -r path/to/requirements.txt
How to run it
-
Download the model from this link (it exceeds 100 megabytes so it can't stay on Github) and place it in a folder called
modelin the main directory of the project. -
Run
predictions.py. The script will return the classification report and the confusion matrix of the model.
Achievements
Actually the model has an F1 score of 91% on test data.
If you want to train again the model changing the parameters, do your changes in train.py and then launch it.
Metrics (Classification Report and Confusion Matrix)
Model structure
