# Evaluation
## Metric
The evaluation measure for this competition is the classification accuracy computed on the test set. The classification accuracy is given by the number of correctly predicted labels divided by the total number of test samples.

## Public versus Private Test Evaluation
The public leaderboard is calculated on approximately 50% of the test data. These examples are randomly chosen. The final results will be based on the other 50%, so the final standings may be different. In this context, you will have to choose 2 submissions that you think will attain the best performance on the 50% of the test data that is not used for the public leaderboard. 

## Submission Format
For every sample in the test set, submission files should contain two columns: Image and Class. The Image is the file name of an image sample. The Class should be the class label, 0 to 95, predicted for the corresponding data sample.

The file should contain a header and have the following format:

Image,Class
16865.png,0
10070.png,0
...