CS623: Multimedia Surveillance Systems
Semester SI, 2023-24

 

Course Information Lectures/Calendar Quizzes Labs

Course Information

Lectures (CS2): Tue - 4:00 PM, Wed - 5:00 PM
Labs (Lab 2): TBA

Objectives

Objective of the course is to give an overview of is to give an overview of statistical and machine learning based ANOMALY DETECTION techniques used for automatic surveillance using multimedia data. The course is designed to provide hands-on experience of the techniques with real-world data.

Outcomes

By taking this course, the students will be able to find answer to the following questions:
  • What is anomaly?
  • How to detect anomaly in a time series data?
  • What are different statistical and machine learning techniques for anomaly detection?

    • Prerequisite

    Multimedia Systems (CS507) OR Digital Image Processing and Analysis (CS517) OR Computer Vision (CS517) OR Image processing and pattern recognition (EE484) OR MAchine Learning OR any Equivalent Course.

    Course Requirements

    Student are required to attend two lectures per week and appear in two exams. In addition, there will be weekly lab sessions. During lab sessions, the students are required to solve and implement programming assignments.

    Grading Policy

    There will be lab exercises, quizzes, a mid-semester exam, a final exam and project. The tentative marks distribution is as follows:
    Quizzes (top 2 out of 3): 10%
    Lab Exercises (top n-1): 20%
    Mid-semester exam: 20%
    Final exam: 20%
    Project: 30%

    Attendance Requirement

    The course will follow institute policy on attendance, i.e., the students should attend at least 75% of the lectures to get non F grade. If your attendance is above 90%, you will get 1 bonus mark. The same policy applies to attendance in the lab sessions.

    Code of Ethics & Professional Responsibility

    It is expected that students who are taking this course will demonstrate a keen interest in learning and not mere fulfilling the requirement towards their degree. Discussions that help the student understand a concept or a problem is encouraged. However, each student must turn in original work. Plagiarism/copying of any form, will be dealt with strict disciplinary action. This involves, copying from the internet, textbooks and any other material for which you do not own the copyright. Copying part of the code will be considered plagiarism. Lending the code to others will be considered plagiarism too, for it is difficult to investigate who copied whose code. Students who violate this policy will directly receive a failing grade in the course. Remember - Your partial submission can fetch you some points, but submitting other's work as your own can result in you failing the course. Please talk to the instructor if you have questions about this policy. All academic integrity issues will be handled in accordance with institute regulations.

    Textbooks

    Primary Textbook

    There is no single textbook for the course. We will rely heavily on the web sources for the content. Few possible reference books are given below:

    Reference Books

    1. Outlier analysis, Authors: Charu C. Aggarwal, Publisher: Springer, Cham, 2015.
    2. Anomaly Detection for Monitoring, Authors: Baron Schwartz and Preetam Jinka, Publisher: O'Reilly Media, Inc., Year 2016. [Link].
    3. Multimedia video-based surveillance systems: Requirements, Issues and Solutions, Authors: Foresti, Gian Luca, Petri Mähönen, and Carlo S. Regazzoni, Publisher: Vol. 573. Springer Science & Business Media, 2012.

    Language/Tools

    For lab exercises we will primarily use Python. For projects, students are free to use any language.

    Teaching Assistant

    Shreya Bansal (shreya.22csz0010@iitrpr.ac.in)

    Contact Me

    By appointment at
    Room No. 319, S. Ramanujan Block, Permanent Campus, IIT Ropar

    Tentative Topics

    • Anomaly modelling using statistical techniques - Z-score, LoF, Clustering, Mixture Models, PCA
    • Anomaly detection using Machine Learning Techniques - One Class SVM,GANs, Auto-encoders, Transformers
    • Anomaly detection in time-series data - ARMA/ARIMA, AGMM, HMM, DTW, LSTM
    • Reconstruction and memory based methods, Multimodal methods, Compression-based methods
    • Case Studies

    Quizzes

    Quiz 1 - TBA
    Quiz 2 - TBA
    Quiz 3 - TBA

    Projects

    Projects are to be done individually or in a group of two. The end application of the project should be in one of the following application areas: Agriculture, Surveillance, Health, Education, Agriculture, Sports, Public Transportation. Project requirements:
    • The reports must be prepafed in ACM Multimedia LaTeX format. Good quality English is expected in the report.
    • The code should be submitted through GitHub or Bitbucket repository. You can make a private repository and show me with your login. I will observe the activities on repository (commits, etc.) to check the progress.
    • Dataset can be submitted through Pen Drive of Google Drive.
    • You are free to use resources (code) available on the Internet with proper references. However, during evaluation you need to explicitly mention parts with your work.
    • There will be marks for creativity in the project.
    • There will be 4 to 5 evaluations of the project.

    Lab Exercises

    Most lab sessions will be guided, with an unguided component need to be submitted within next 2 days.

    Lectures and Calendar

    Lectures Dates Topics Readings Events
    L1 July 25 Introduction

    *This is a tentative schedule. The schedule can change according to the need at the discretion of the instructor.