Computer Vision Methods Labs

Schedule (CZ course) Schedule (EN course) Upload system Discussion forum

Labs plan

Labs are organized in four main topics: Correspondence problem, Indexing and image retrieval, Object tracking, Convolutional neural networks. Each topic is covered by approximately two to four labs.

Week	Date	Topic	Teacher	Recording
1	18.2.	Introduction to Image Processing in python using PyTorch.	DM	recording 2022
2	25.2.	Debugging pytorch code.	DM
3	4.3.	Correspondence problem I, detection of the interest points.	DM	recording 2022
4	11.3.	Correspondence problem II, computing local invariant description.	DM	recording 2022
5	18.3.	Correspondence problem III, finding tenative correspondences and RANSAC.	DM	recording 2022
6	25.3.	Correspondence problem, summary.	DM
7	1.4.	Convolutional Neural Networks: training a classifier	LN	recording 2022
8	8.4.	Convolutional Neural Networks II: debugging training process	LN
9	15.4.	Image Retrieval, BoW TF-IDF, fast spatial verification.	PS, GKZ
10	22.4.	Assignment defence
11	29.4.	Deep metric learning	PS, GKZ
12	6.5.	Self-supervised Learning	PS, GKZ
13	13.5.	Rector's sport day (no teaching at the CTU)
14	20.5.	Tracking I, Kanade-Lucas-Tomasi tracking (KLT tracker)	JS

Labs will be accompanied with a simple programming task. Detailed specification of the tasks is described in each of the labs. Students will upload their results and their codes through the BRUTE system.

Each lab will usually consists of three parts:

Discussion on the last lecture. Students will be free to ask any questions related to the last lecture. At the end of this session, a teacher will pose a question to a volunteer/random student. If the student answers correctly, he/she will get a single bonus point.
Working on the current task. Students are free to ask any specific questions, discuss their current results, resolve any programming issues.
Short introduction of the next task. The teacher will briefly introduce the next problem, give some hints and answer possible questions.

Assessment

You are obliged to carry out all programming tasks at least a minimal required quality. All tasks must be carried out individually, with a possible help of AI. You are free to discuss the problems with your colleagues, however the code must be written strictly by yourself. See plagiarism if you are unsure what is allowed.

AI usage policy

You are free to use tools like ChatGPT to help you with the assignment. During the labs, you would be required to answer questions about the code and explain how it works, how it could be modified, what you are proud of, what you have struggled with, etc.

Such defence of the assignment will be done several times per semester and will give you a coefficient from 0 to 1.2, which then is multiplied by score in BRUTE. This way one could have perfect assignment in BRUTE, but if he or she is not able to explain how it works, the final score for the lab will be zero.

Evaluation Policy

The points from the labs will contribute to 50 percent of your course evaluation.

There will be 11 tasks awarded with points throughout the semester (a new task will be given out every week in the lab, with the exception of two labs which are intended to help students with debugging). On top of these tasks, there will also be ability to get bonus points in some labs.

Each task has a deadline of 2 weeks and 1 day.
This is mostly relevant to the defence of the lab (see the AI policy above)

Useful Links

Computer vision: Facts & Fiction series
Computer Vision course at CMU
PyTorch & Python development

Questions

If you have any question, use our discussion forum preferably. For organizational questions, contact Jan Čech, who is responsible for the MPV labs. For technical questions, please contact the relevant teaching assistant.

Course Assistants

Lukáš Neumann	Dmytro Mishkin	Pavel Šuma	Jonáš Šerých	Giorgos Kordopatis-Zilos	Jan Čech (lead)

Table of Contents