Project 1 - Grading Scale
-
Change glickman recognizer to sunglass recognizer - 40 pts
-
Evidence of correct code changes - 30 pts
-
Questions about the network - 10 pts
-
What code did you modify? - 2 pts
-
What was the maximum classification accuracy achieved on training? - 2pts
-
How many epocs did it take to reach this level on training? - 2pts
-
What was the maximum classification accuracy achieved on the validation
set? - 1pt
-
How many epocs did it take to reach this level on validation? - 1pt
-
What was the maximum classification accuracy achieved on test set? - 1pt
-
How many epocs did it take to reach this level on the test set? - 1pt
-
Changes for a 1-of-20 face recognizer - 30 pts
-
Code changes - 10 pts
-
Output encoding and explanation - 5 pts
-
Questions about the network - 5 pts
-
What code was modified? - 1 pts
-
What was the maximum classification accuracy achieved on the training set?
- 1 pts
-
How many epocs did it take to reach this level on training? 1 pts
-
Classification accuracy and epocs on validation set - 1 pt
-
Classification accuracy and epocs on test set - 1pt
-
Commenting on the commonality between misclassified images - 10 pts
-
Changes for a pose recognizer - 30 pts
-
Code changes - 10 pts
-
Out encoding and explanation - 5 pts
-
What was maximum classification accuracy on training set? - 1 pt
-
How many epocs did it take to reach this level? - 1 pt
-
What was maximum classification accuracy on validation set - 1 pt
-
How many epocs did it take to reach this level? - 1 pt
-
What was maximum classification accuracy and epocs on test set? - 1 pt
-
Comment on which regions of the image the hidden units are weighting more
- 5 pts
-
Comment on hidden units tuning to different features of the image - 5 pts
-
Additional Modifications - up to 10 bonus pts
-
Rewarded for outstanding insights on the assignment and/or additional insightful
code changes or well thoughtout comments.