• Description

The Pattern Recognition course will introduce the student to the techniques for extracting information from data. In particular the course focuses on the recognition of patterns within the context of computer vision. Images are one of the main sources of information used by the human brain at the perceptual level to make decisions. As a consequence, the recognition of patterns within the context of computer vision is of great interest, especially nowadays that we have at our disposal a huge amount of visual data that can not be analyzed by hand. Practical applications of pattern recognition in the context of artificial vision are many. For example safety, medicine, automatic inspection, or automatic navigation.

• Prior knowledge

• Requirements: Ability to read scientific texts in English. Basic statistical knowledge (undergraduate or engineering level).
• Bibliography (course material): Richard Szeliski, 2022. "Computer Vision: Algorithms and Applications", Second Edition.

• Information prior to enrolment

• Coordinator: Dr. Carles Ventura
• Credits: 6
• Description: In the present course the student will be introduced in the set of techniques that allow us to extract information from a set of data. As an application, the subject will focus on extracting high level information from the environment, from information captured by cameras. In this case the objective is to learn to recognize objects in real environments and completely automatically. Images are one of the most important sources of information that the human brain uses perceptually to make decisions. In this context, different samples of the same object have in common a series of patterns, which must be detected, modeled and later classified for recognition. The course aims to train the student to know the probabilistic techniques and mathematical optimization related to pattern recognition.
• Requirements: ability to read scientific texts in English. Basic statistical knowledge (undergraduate or engineering level).
• Bibliography: UOC modules

• Learning objectives and results

This course is oriented to the study of fundamental concepts of computer vision, pattern recognition and other advanced topics related to the problems of analysis and automatic recognition of complex images. In particular, the learning objectives are as follows:
 
- To know how the images are formed.
- To know the main techniques of digital image processing.
- To understand how color is perceived and know the spaces of color representation.
- To know the main techniques of dimensionality reduction (selection and extraction of characteristics), both supervised and unsupervised, and know how to apply to real problems.
- To know the main techniques of automatic learning for automatic classification of data and know how to apply to real problems.

The specific competences that are covered in this course are the following:
 
A1. Ability to understand and be able to apply advanced computer knowledge and numerical or computational methods to engineering problems.

A2. Ability to apply computational, mathematical and statistical methods to model, design and develop applications, services, intelligent systems and / or systems based on knowledge.

A3. Ability to apply mathematical and computational methods to solving technological and engineering problems, particularly in research, development and innovation.

A4. Ability to model problems using a mathematical language and solve them through formal reasoning.

A5. Ability to identify mathematical theories necessary for the construction of models from problems of other disciplines.

A6. Ability to handle mathematical and statistical software.

A9. Ability to analyze and process data to generate and manage useful information in decision making.

A10. Ability to design, implement and validate algorithms using the most convenient structures.

• Content

Unit 1. Introduction to computer vision. A brief history of computer vision. Color spaces.

Unit 2. Image processing. Linear filters and non linear filters. Morphological operations.

Unit 3. Feature extraction. Feature detection and description. Feature matching. Content-based image retrieval. Segmentation.

Unit 4. Machine Learning. Supervised and unsupervised learning. Deep Neural Networks. Convolutional Neural Networks.

Unit 5. Image Classification. Feature-based methods. Convolutional Neural Networks based methods.

Unit 6. Object detection. Face detection. Person detection. Object detection.

Unit 7. Semantic segmentation. Instance segmentation. Panoptic segmentation. Pose estimation.

Unit 8. Video understanding. Action recognition. Optical flow. Object tracking. Video object segmentation.

Tema 9. Generative models. Autoencoders. Variational autoencoders. Generative Adversarial Networks (GANs).

• View the UOC learning resources used in the subject

• Additional information on support tools and learning resources

For the practical exercises, the resources than can be found on the following code repository will be used:

https://gitlab.uoclabs.uoc.es/patternrecognition/pattern-recognition/-/tree/master

• Guidelines on assessment at the UOC

Assessment at the UOC is, in general, online, structured around the continuous assessment activities, the final assessment tests and exams, and the programme's final project.

Assessment activities and tests can be written texts and/or video recordings, use random questions, and synchronous or asynchronous oral tests, etc., as decided by each teaching team. The final project marks the end of the learning process and consists of an original and tutored piece of work to demonstrate that students have acquired the competencies worked on during the programme.

To verify students' identity and authorship in the assessment tests, the UOC reserves the right to use identity recognition and plagiarism detection systems. For these purposes, the UOC may make video recordings or use supervision methods or techniques while students carry out any of their academic activities.

The UOC may also require students to use electronic devices (microphones, webcams or other tools) or specific software during assessments. It is the student's responsibility to ensure that these devices work properly.

The assessment process is based on students' individual efforts, and the assumption that the student is the author of the work submitted for academic activities and that this work is original. The UOC's website on academic integrity and plagiarism has more information on this.

Submitting work that is not one's own or not original for assessment tests; copying or plagiarism; impersonation; accepting or obtaining any assignments, whether for compensation or otherwise; collaboration, cover-up or encouragement to copy; and using materials, software or devices not authorized in the course plan or instructions for the activity, including artificial intelligence and machine translation, among others, are examples of misconduct in assessments that may have serious academic and disciplinary consequences.

If students are found to be engaging in any such misconduct, they may receive a Fail (D/0) for the graded activities in the course plan (including final tests) or for the final grade for the course. This could be because they have used unauthorized materials, software or devices (such as artificial intelligence when it is not permitted, social media or internet search engines) during the tests; copied fragments of text from an external source (the internet, notes, books, articles, other students' work or tests, etc.) without the corresponding citation; purchased or sold assignments, or undertaken any other form of misconduct.

Likewise and in accordance with the UOC's academic regulations, misconduct during assessment may also be grounds for disciplinary proceedings and, where appropriate, the corresponding disciplinary measures, as established in the regulations governing the UOC community (Normativa de convivència).

In its assessment process, the UOC reserves the right to:

• Ask students to provide proof of their identity as established in the UOC's academic regulations.
• Ask students to prove the authorship of their work throughout the assessment process, in both continuous and final assessments, through a synchronous oral interview, of which a video recording or any other type of recording established by the UOC may be made. These methods seek to ensure verification of the student's identity, and their knowledge and competencies. If it is not possible to ensure the student's authorship, they may receive a D grade in the case of continuous assessment or a Fail grade in the case of the final assessment.

Artificial intelligence in assessments

The UOC understands the value and potential of artificial intelligence (AI) in education, but it also understands the risks involved if it is not used ethically, critically and responsibly. So, in each assessment activity, students will be told which AI tools and resources can be used and under what conditions. In turn, students must agree to follow the guidelines set by the UOC when it comes to completing the assessment activities and citing the tools used. Specifically, they must identify any texts or images generated by AI systems and they must not present them as their own work.

In terms of using AI, or not, to complete an activity, the instructions for assessment activities indicate the restrictions on the use of these tools. Bear in mind that using them inappropriately, such as using them in activities where they are not allowed or not citing them in activities where they are, may be considered misconduct. If in doubt, we recommend getting in touch with the course instructor and asking them before you submit your work.

• View the assessment model