Fall 2019 Assignments
For all assignments, please create a page in your own wiki space and give it the label provided at the end of the assignment. You may work in groups of up to three people.
Assignment 1: Basic Vision Task
Due 25 September 2019
Pick a basic visual perception task like detection, classification, tracking, or measurement from a single image. Keep it simple. Implement a solution that makes use of a deep network. Try at least two variants (discuss your choices with the professor). Variants may include variations in the pre-processing steps rather than different networks. You may pick an existing image database for training or build your own. Your report should describe your design, data preparation, the two variants, and an evaluation of the results using standard metrics.
Handin label: cs465f19project1
Assignment 2: Speech Recognition
Due 18 October 2019
Pick a speech detection/recognition task. Keep it simple, such as isolated word recognition, and make use of an existing annotated database. Implement a solution that makes use of a deep network, possibly recurrent. Try at least two variants of pre-processing the data and/or network architectures. Your report should describe your design, data preparation, the two variants, and an evaluation of the results using standard metrics.
Handin label: cs465f19project2
Assignment 3: Video
Due 13 November 2019
Pick a video analysis task. Keep it simple, such as single action detection. You may make use of existing databases or collect and annotate your own. Only do the latter if it will also be used for your final project. Implement a solution that makes use of a deep network, possibly recurrent. Try at least two variants of your solution. A variation may involve pre-processing or it may involve different network architectures. Your report should describe your design, data preparation, the two variants, and an evaluation of the results using standard metrics.
Handin label: cs465f19project3
Assignment 4: Final Project
Due at scheduled final exam
-
In consultation with the professor, select a final project topic and implement it. Any aspect of perception is fair game for this project. You can select a technique from a paper and implement it, or you can pick something new and build it from scratch. Whatever you do, keep it focused. You do need to demonstrate the performance of your system using some kind of standard metric.
Handin label: cs465f19project4