This tool chain includes a complete face recognition experiment for face recognition algorithms that do not touch the training set. The tool chain is aligned such that the columns are grouped for the different steps in face recognition, which are as follows:

1. Database: The database is split into two sets, for the development (dev) and test set; it provides images and hand-labeled eye locations; note that the eye locations are in subject perspective (i.e., the right eye has a lower x coordinate than the left eye.
2. Image alignment: The raw images from the database are aligned accoring to their hand-labeled eye locations, and the aligned image is returned.
3. Preprocessing: A preprocessing such as a photometric enhancement is applied to the aligned images.
4. Feature extraction: Features are extracted from the preprocessed images. The feature extraction stage cannot have a training stage, but all required data needs to be specified as parameters.
5. Template enrollment: Templates are enrolled from one or more features of images from the same identity.
6. Scoring: In the scoring stage, templates and probe features are compared, and for each pair, a score is produced.
7. Sorting: The scores are split into positive and negative scores, according to the ground truth information of the identites in the database.
8. Evaluation: The evaluation of the scores is split into the evaluation of both sets indiviually (where, i.e., ROC curves can be plotted) and a combined evaluation, where a threshold can be selected on the development set, and applied to the test set.