Grading

The grading breakdown for this class is as follows:

Component Weight
Five self-graded homeworks 25%
Midterm 1: classical vision 25%
Midterm 2: modern vision 25%
Final project 25%

Homework in this course is primarily formative: it is where you practice the material and learn from your mistakes. The two midterms verify individual mastery, and the project evaluates your ability to carry out research.

The cutoffs for letter grades vary each year, but here are the cutoffs we used in a previous offering of the class: A (92), A- (87), B+ (83), B (79), B- (75), C+ (70), C (65), C- (60), D+ (55), D (50).

Midterms

There will be two midterm exams for this class, both held in class:

  • Midterm 1: classical vision — 2:30–3:45 PM, Thursday, October 15, Computer Science Building W150
  • Midterm 2: modern vision — 2:30–3:45 PM, Tuesday, December 8, Computer Science Building W150

Midterm 1 covers the classical vision material through optical flow and point tracking; Midterm 2 covers the modern vision material from learned representations onward. Both are taken individually and without AI assistance.

Makeup exams: Students are expected to take the exams at the scheduled times unless there are extenuating circumstances. We will follow the University guidelines for attendance when determining eligibility for makeup exams.

Homework Assignments

There will be 5 homework assignments over the course of the semester, each worth approximately 5% of your final grade. These assignments may contain material that has been covered by published papers and webpages.

Guided Self-Assessment

Homework in this course is graded through guided self-assessment. The purpose is to practice the material, identify gaps in your understanding, and learn to evaluate your own technical work accurately.

Each assignment has two stages:

  1. Original submission. Submit your work before the original-submission deadline. You should make a serious attempt at every problem and show enough reasoning, code, and intermediate results for your approach to be evaluated. Solutions will be released to the class after this deadline.

  2. Self-assessment. After solutions and a detailed rubric are released, submit:

    • a diagnostic correctness score for each problem and for the assignment as a whole;
    • corrections to any substantive errors in your original work;
    • a brief explanation of what went wrong and how you corrected it; and
    • a proposed homework grade based on the process rubric below.

The diagnostic correctness score is intended to help you assess your mastery, but it does not directly determine your homework grade.

Your homework grade is based on the following process rubric:

Component Weight
Serious and complete original attempt 50%
Accurate and thoughtful self-assessment 50%

A serious original attempt addresses every problem, explains the approach taken, and reflects your own understanding. An accurate and thoughtful self-assessment identifies errors honestly, applies the rubric correctly, and explains the source of each substantive mistake.

You are not penalized for reaching an incorrect answer in your original submission. You may, however, lose credit for an incomplete or superficial attempt, an answer you cannot explain, inaccurate self-scoring, incomplete corrections, or an error analysis that does not engage with what went wrong.

Collaboration and AI Policy

Homework is intended for practice, so discussion and responsible tool use are allowed. However, every submission must reflect your own work and understanding.

  • You may discuss concepts with classmates. You may help one another understand the material, compare high-level strategies, and debug conceptual or implementation problems.
  • You must prepare your submission independently. Your code, derivations, figures, and explanations must be written by you. Do not copy, jointly write, share, or view another student’s completed solution before the original-submission deadline.
  • You may use AI assistants and other tools. However, you remain responsible for checking their output and understanding everything you submit. Using an AI assistant to produce an answer does not by itself constitute a serious attempt if you cannot explain or evaluate that answer.
  • You must disclose collaboration and AI assistance. On each submission, list the people with whom you worked. For AI use, identify the problems on which it was used and briefly describe how it contributed—for example, explaining a concept, debugging code, suggesting an approach, or revising prose. A complete transcript is not required unless the assignment says otherwise.
  • Self-assessment must reflect your own judgment. You may seek clarification from classmates, course staff, or AI tools, but your scoring, corrections, and explanations of what went wrong must be written independently and must accurately represent your understanding.
  • Project collaboration is limited to your project group. Assistance from people outside the group and substantive use of AI-generated code, text, images, or experimental analysis must be disclosed. Additional requirements may be specified on the project page.

Copying completed work, sharing solutions before the deadline, submitting work you do not understand, or failing to disclose substantive assistance may constitute a violation of the University’s Academic Integrity Policy.

Deadlines and Late Work

Because solutions are released immediately, late original submissions cannot ordinarily be accepted. When an approved accommodation or documented extenuating circumstance applies, the instructor may provide an alternate assignment or reweight the affected homework; contact the instructor as soon as reasonably possible.

Regrade Policy

For the homework, midterms and the project, submit a regrade request on Gradescope within the specified deadline. Please note that a regrade may cause your score to go either up or down.

Project

The project will be carried out in groups of two or more people, and has three main parts: a proposal, a final report, and a poster/oral presentation. The project is an integral part of this class, and is designed to be as similar as possible to researching and writing a conference-style paper. Please see the project page for more information.

Textbooks

The primary material for the class are lectures and readings from books, research papers, articles listed on the lectures page. There is no required textbook for this class. Nevertheless the following textbooks are useful:

We will post links to sections of Szeliski’s book for each lecture.

And these for machine learning:

Required Background

The course assumes a strong ability to program in Python and background in linear algebra, probability and statistics. Take a look at the resources below to brush up your math and programming skills.

Writing is a key component of all projects. Here are some tips on how to write well.

Past offerings of 670 at the university

Related courses at the university

  • 682: Neural Networks: A Modern Introduction
  • 590: Intelligent Visual Computing
  • 373: Introduction to Computer Graphics
  • 370: Introduction to Computer Vision

Accommodation Statement

The University is committed to providing an equal educational opportunity for all students. If you have a documented physical, psychological, or learning disability on file with Disability Services (DS), you may be eligible for reasonable academic accommodations to help you succeed in this course. If you have a documented disability that requires an accommodation, please notify me within the first two weeks of the semester so that we may make appropriate arrangements.

Academic Honesty

Since the integrity of the academic enterprise of any institution of higher education requires honesty in scholarship and research, academic honesty is required of all students. Students are expected to be familiar with this policy and the commonly accepted standards of academic integrity (http://www.umass.edu/honesty).

Acknowledgements

Many of the slides and homework assignments are based on excellent computer vision courses taught elsewhere by Svetlana Lazebnik, Alyosha Efros, Alexander Berg, Steven Seitz, James Hays, Charless Fowlkes, Kirsten Grauman and many others. Many thanks to Richard Szeliski for making the computer vision textbook available online for free.