Each directory in this repository includes a submission to the Kaggle competition Yelp Restaurant Photo Classification.
-
/CNN_Submission1: This starter code uses CaffeNet to extract image features, takes the mean of image features to represent restaurants, and uses SVM to classify restaurants. Private LB F1-score = 0.76094. (More Discussion)
-
/HOG_Submission1: Benedict's submission; similar to CNN_Submission but uses histogram of oriented gradients (HOG) features.
-
/BagDistance_Submission1: This computes the Chamfer distances between restaurants and uses SVM classifiers. Private LB F1-score = 0.82219