Pattern Recognition

(CSCI-737-01, Spring 2020)
Lecture: Tues./Thur. 2-3:15pm (GOL 1610)
Instructor: Dr. Richard Zanibbi (
Office Hours: Wed 3-4:30pm, Thurs 3:30-5pm (GOL 3551)


Final Exam

  • **UPDATE** For the bonus, submit regular label graph files (e.g., in OR format). We do not need to use symLG, because no one is using a system that cannot produce stroke-level graphs directly.
  • UPDATE: for your non-baseline parser, go ahead and use whatever input format the parser already uses (e.g., for SESHAT or LPGA). 
  • **IMPORTANT** make sure to consult the lecture notes and papers posted for Week 15 - you will find notes on parsing (and evaluating parse results), along with papers with associated math parser implementations for use/discussion in the exam.
  • Our final exam will be held Friday May 1st, from 1:30-4pm over Zoom (a link is provided under Zoom" on MyCourses). This will be an in-class presentation for Project 3, worth 10% of your final grade, and the project 3 submission will be worth 15% of your grade (25% in total for Project 3).
  • Final Project 3 materials will be due after the exam, on Tuesday May 5th at 11:59pm (late submissions will not be accepted, due to time constraints).
  • Exam Presentations: Presentations will be given by each Project 3 group on Friday May 1st. Talks will be 10 minutes long, with 5 minutes for questions. The talk should include the following: 1) (quickly, 1 slide) results and summary of errors for the baseline parser (see project write-up); 2) overview of an existing math parser that you have installed and run, including key algorithms and example inputs and outputs, 3) a discussion of how you will modify this system, plans for using a different system, or a sketch for a system that you will implement (preliminary results welcome, but not required). Presentations should include citations to research papers associated with the systems/techniques that you discuss. Include these citations on the slides where you discuss them; do not put them all on a slide at the end of the talk. 
  • Please used the updated versions for CROHMELib and training Data on MyCourses for Project 3. This resolves problems with multiple relationships for 'Inside' relationships (in the script and data), and provides a copy of the newest InkML web-based visualization tool (from CROHME 2019).

Week 15

  • Thursday will be our last class for the semester.  Thank you all for your continued attendance in class despite all the challenges facing us at the moment.
  • Please complete the on-line course evaluation. I rely upon student feedback to improve the course for future offerings; I'd particular appreciate hearing about what you felt worked, and what could be improved in the future.
  • Office Hours: Prof. Zanibbi will hold regular office hours this week and next week (Wed/Thurs until May 1st), and will continue to respond to email and the public discussion forum on MyCourses within a day. 
  • Congratulations to our Project 2 bonus winners: 1) Yifei and Shuo, 2) Nikhil and Rahul, and 3) Shaun and Paritosh. These three groups placed 1st-2nd-3rd for both the symbol segmentation and segmentation + classification tasks.
  • All project 2 reports are posted on MyCourses (with the project). All presentation slides given during the semester have also been posted with their associated research papers.
  • **Extension the 2nd.** Project 2 is due Tuesday April 21st at 11:59pm.
  • We will talk about parsing handwritten mathematical notation and the exam during this last week of class.  

Week 14

  • **Extension the 2nd.** Project 2 is now due Tuesday April 21st at 11:59pm. A3 will be graded ASAP.
  • Class this Thursday, we'll meet to discuss Project 2. I will not be presenting any new material until next week (when we'll talk about parsing, and I'll share some related work).
  • There will be no more paper presentations for the semester, and there will be no 4th assignment.  The third paper presentation will be given as a bonus to everyone.
  • Exam update: we will have our final project presentations and submissions as planned, but with an increased weight of 25%: 10% for the in-class presentation ('exam'), and 15% for the project submission.
  • Please complete the on-line course evaluation when you can. I need your feedback to  improve the course for future students. Knowing both what you would keep in the course and what you would change would be very helpful. 

Week 13

  • *Extension the last*: Assignment 3 is due Thursday (Apr. 9) at 11:59pm. Due to personal matters, the grading with Project 1 is delayed; we are moving the A3 deadline up accordingly. I plan to grade the assignments ASAP.
  • Reading: Please read the SSD.pdf paper on MyCourses, on the Single Shot Detector. We will discuss this in class later this week.

Week 12

  • Papers for presentation on Thursday are on MyCourses. Group 4: FocalLossObjectDetection, Group 5: ObjectTrackingClustering, Group 6: TrainingDeepDetectorFromScratch.
  • *Extension the last*: Assignment 3 is due Thursday (Apr. 9) at 11:59pm. Due to personal matters, the grading with Project 1 is delayed; we are moving the A3 deadline up accordingly. I plan to grade the assignments ASAP.
  • **Extension.** Project 2 has been posted on MyCourses. It is due April 14 at 11:59pm.
  • To help with running your experiments for the next project, the storage on your CS accounts increased to 25GB, and scikitlearn has been updated on the CS network machines. For anyone wanting to use GPUs for Project 2, students in the class now have access to the GPU servers,, and Each have 10 GPUs apiece - but at least two other classes are using these machines; you should use only one GPU for your experiments. Use the command 'nvidia-smi' to check on card utilization, be aware of which specific card you are using, and avoid conflicts with other students. 
  • Reading for this week: Casey & Lecolinet paper on character segmentation (under 'Sources' on MyCourses).

Week 11

  • On Thursday, Groups 1-3 will present papers on segmenting mathematical symbols in formulas during lecture (over Zoom - link is available from MyCourses -> Zoom). Papers are available through MyCourses. Paper Assignment: Group 1: HMM-Seg+Rec, Group 2: MathSymbolSegRec-Lattice, Group 3: icfhr2016_los_final. 
  • Assignment 3 has been posted. It is due next Friday at 11:59pm. Project 2 will be assigned soon as well (it is related to Assignment 3).

Weeks 9-10

  • Update: due to the coranavirus, the course will be delivered on-line for the remainder of Spring 2020. Lectures and meetings during office hours will be held over Zoom; there is also a new discussion thread for students to ask questions publicly anytime. Students should feel welcome to send me email if they prefer to speak privately. Details are provided on MyCourses under Content -> Administration.
  • The course schedule has been updated. The course syllabus remains unchanged.  
  • Course deliverables will remain the same. A few changes/notes:
    • Paper presentations will be given during lecture online, by groups using Zoom. Presentation slides should be uploaded to the appropriate assignment box.
    • The deadlines for deliverables have been moved up (see the course schedule)
    • The exam (i.e., final project presentation) will be held at the original time, but the final project will be due two days later (Sunday May 3 at 11:59pm).
    • Projects will be compelted as planned by teams of two students. You are strongly encouraged to collaborate over Zoom as well as email/source code repositories.

Week 8

  • **Extension (2nd):** Project 1 is due Sunday (Mar 8) at 11:59pm. The submissions web site for the bonus is located here.
  • We will finish classification this week, and then move on to segmentation.

Week 7

  • **Extension:** Project 1 is due Friday (March 6) at 11:59pm. A submission website for the bonus is now online. This link is also available through MyCourses under "Project 1."
  • Due to illness, office hours and lecture on Thursday are cancelled this week.
  • Grades for A2 will be available by the early part of next week.

Week 6

  • **Extension #2:** Assignment 2 is due Tuesday@11:59pm (hard deadline). This is due to an unexpected quirk in a new grading feature on MyCourses.
  • Groups 4-6 will give paper presentations in-class on Thursday. Paper assignments and details are provided below.
  • Students should have now received feedback on Assignment 1, both as annotations on their submitted write-up, as well as general comments to the class.
  • Project 1 is due Sun., March 1st at 11pm. A submission website for the bonus is now online. This link is also available through MyCourses under "Project 1."

Week 5

  • **Extension**: Assignment 2 is due Sunday@11:59pm. The write-up is available on MyCourses.
  • Project 1 will be posted soon on MyCourses, and is due Sunday March 1st at 11pm.
  • Papers for presentation next Thursday in-class are on MyCourses.  Paper assignments: Grp 4: PreprocessGen, Grp5: HybridFeatures, Grp 6: SpatialRelationships. Please note that these papers are shorter - they are conference papers related to the course project. We will switch which groups present TPAMI vs. project-related papers in the next round. 

Week 4

  • Assignment 2 is due next Friday (Feb. 14).  The assignment write-up is provided on MyCourses, and concerns using estimated Bayesian classifiers with 2D Gaussian feature distributions for each class.
  • Paper presentations from Groups 1-3 will be given in-class this Thursday.  Paper assignments can be found below, and other requirements are detailed on MyCourses. For those groups presenting, make sure to submit your slides to the dropbox/assignment box before class.

Week 3

  • Assignment 1 is due Friday Jan. 31st at 5pm.
  • Paper Presentations next Thursday: Grp 1: 'Atomic.pdf,' Grp 2: 'Hyperplanes.pdf,' Grp. 3: 'LossFunctions.pdf.' See MyCourses to find your group members (under 'Groups') and presentation requirements.

Week 2

  • Assignment 1 has been posted on MyCourses. It is due Friday Jan. 31st at 5pm. Writing and programming are required for the assignment.
  • Make sure to read Ch. 2 of Hastie; we will be moving on to reading the chapter from Duin et al. on Bayesian Decision Theory next (available through MyCourses).

Week 1

This is the home page for the Pattern Recognition course being offered by the RIT Department of Computer Science in Spring 2020. 
  • The course syllabus and schedule are under construction.
  • Readings are listed on the course schedule, and sources will be made available through MyCourses (there is no course textbook).
  • The course will include assigned readings, assignments, and (team) projects. You will be expected to analyze (e.g., when 'math-ing'), program/test/document, and read and write thoughtfully for this course. Students will also be reading and discussing research papers.
  • Please note my updated office hours (see above).
  • The course exam has been set (the 'exam' will be Project 3 presentations in our regular classroom (GOL 1610)).