Advanced Machine Learning for Computer Vision

Course
2021-2022
Semester
2
ECTS
6
Type
Compulsory
University
UVigo and UDC

Subject objectives

The objective of this subject is to know and apply advanced neural models, to know the techniques of the state of the art of deep learning, with end-to-end training approaches, and minimizing the use of tagged data, to solve computer vision applications using the methodologies covered in the subject.
-To know, apply and evaluate advanced neural models.
-To know deep learning techniques, with end-to-end training approaches, and minimizing the use of tagged data.
-To solve computer vision applications using advanced machine learning methods.

Contents

Multilayer perception and backpropagation.
Convolutional and recurrent networks
Principles of deep learning
Self-supervised learning and autoencoders
Advanced neural models for computer vision.
Advanced supervised learning paradigms
Selected topics in machine learning for computer vision
Advanced applications in computer vision.

Basic and complementary bibliography

Basic:
Recent papers from relevant scientific journals and conferences: NIPS, ICML, IJCAI, AAAI, ECML, CVPR, ICDM, IEEE PAMI, IEEE TKDE, etc.

Complementary:
Ian Goodfellow, Yoshua Bengio, Aaron Courville. Deep Learning. MIT Press.
2017.

Competencies

A2 CE2 – To know and apply machine learning and pattern recognition techniques applied to computer vision

B1 CB6 – To possess and understand knowledge that provides a basis or opportunity to be original in the development and/or application of ideas, often in a research contextB2 CB7 – That students are able to apply their acquired knowledge and problem-solving skills innew or unfamiliar environments within broader (or multidisciplinary) contexts related to their area of study
B5 CB10 – That students possess the learning skills to enable them to continue studying in a largely self-directed or autonomous manner

B6 CG1 – Ability to analyze and synthesize knowledge
B8 CG3 – Ability to develop computer vision systems depending on existing needs and apply the most appropriate technological tools
B10 CG5 – Ability to identify unsolved problems and provide innovative solutions
B11 CG6 – Ability to identify theoretical results or new technologies with innovative potential and convert them into products and services useful to society

C1 CT1 – Practice the profession with a clear awareness of its human, economic, legal and ethical dimensions and with a clear commitment to quality and continuous improvement
C2 CT2 – Ability to work as a team, organize and plan

Teaching methodology

Guest lecture / keynote speech:
Participatory lessons with the aim of learning the theoretical content of the subject

Case study:
Elaboration and presentation of selected state-of-the-art methodologies related to the subject.

Objective test:
Continuous evaluation tests during the course.
Evaluation by examination at the end of the course as an alternative.

Laboratory practice:
Analysis and resolution of practical cases with the aim of strengthening the practical application of the theoretical content. Practice in computer classrooms, learning based on the resolution of practical cases, autonomous work and independent study of the students, and group
work and cooperative learning.

Research (Research project):
Learning based on the resolution of practical cases, autonomous work and independent study of the students, and group work and cooperative
learning.

Evaluation system

-Research (Research project) (20):
Resolution of practical cases of application of the subject through autonomous work of the student, and using the techniques learned during the course.
Competences: A2 B1 B2 B5
B6 B8 B10 B11
C2 C1

– Case study (15):
Elaboration and presentation of works on selected state-of-the-art methodologies
Competences:A2 B1 B2 B5
B6 B8 B10 B11
C1 C2

-Laboratory practice (40):
Analysis and resolution of practical cases with the aim of strengthening the practical application of theoretical content.
Competences: A2 B1 B2 B5
B6 B8 B10 B11
C1 C2

-Objective test (25):
Continuous evaluation tests during the course. Evaluation by examination at the end of the course as an alternative.
Competences: A2 B1 B2 B5
B6 B8 B10 B11
C1 C2

Studying time and personal work

Recommended study time for students is about 2 hours per week. Additionally, we estimate that they should spend about 6,5 hours / week working in a number of assignments. All of these activities add up to around 120h/semester.

Subject study recommendations

Subjects that it is recommended to have taken before:
Fundamentals of Machine Learning for Computer Vision /614535007
Image Description and Modeling/614535004

Subjects that are recommended to be taken simultaneously:
Visual Recognition/614535005

Observations

Contingency Plan:

1. Modifications to the contents No change
2. Methodologies All activities are maintained. The teaching will be online and the lessons will take place synchronously in the official
schedule of classes. It may be that, for reasons of inconvenience, some of the classes will be held asynchronously, which will be communicated to the students in advance.
3.Mechanisms for personalized attention to students The tutorials will be telematic and will require an appointment.
4. Modifications in the evaluation No change in the evaluation.
Evaluation activities that cannot be carried out in person will be carried out telematically through the institutional tools in Office 365 and Moodle. In this case, a series of validation measures will be required, which will require the students to have a device with a microphone and a camera, while appropriate validation software is not available. An interview may be arranged with each student to comment on or explain part or all of the tests carried out. In these scenarios, some of the activities under each heading may be modified, adapting them to the situation, but not their overall contribution to the final grade (the weighting percentage)
5. Modifications to the bibliography or webgraphy No change