Flickr classifier

Case Study

This project aims, through machine learning techniques, at creation a model for image classifications, dowloaded by Flickr API. There have been several types of classifications, on a dataset containing images of category "bird" and "mammal". The purpose is to train a model which classifies images as "animal-volatile" or "non-volatile".
The different classifiers are compared with corresponding values ​​of "accuracy_score" that measure the efficiency. In particular have been implemented and compared the following classifiers: 1-nn, 3-nn, 5-nn, Logistic Regression (one-vs-rest), Logistic Regression (multinomial), Gaussian and Multinomial Naive Bayes.

  • Job : Machine learning, Python
  • Date : Jan 2017
  • Agency : University of Catania