Tutorials

Wearable devices equipped with a camera and computing abilities are attracting the attention of both the market and the society, with commercial devices more and more available and many companies announcing the upcoming release of new devices. The main appeal of wearable devices is due to their mobility and to their ability to enable user-machine interaction through Augmented Reality. Due to these characteristics, wearable devices provide an ideal platform to develop intelligent assistants able to assist humans and augment their abilities, for which Artificial Intelligence and Computer Vision play a major role.

Differently from classic computer vision (the so called “third person vision”), which analyses images collected from a static point of view, first person (egocentric) vision assume that images are collected from the point of view of the user, which gives privileged information on the user’s activities and the way they perceive and interact with the world. Indeed, the visual data acquired with wearable cameras usually provides useful information about the users, their intentions, and how they interact with the world.

This tutorial will discuss the challenges and opportunities offered by first person (egocentric) vision, covering the historical background and seminal works, presenting the main technological tools and building blocks, and discussing applications.

Keywords
Wearable, first person vision, egocentric vision, augmented reality, visual localization, action recognition, action anticipation, human-object interaction.

Aims and Learning Objectives
The participants will understand the main advantages of first person (egocentric) vision over third person vision to analyze the user’s behavior, build personalized applications and predict future events. Specifically, the participants will learn about: 1) the main differences between third person and first person (egocentric) vision, including the way in which the data is collected and processed, 2) the devices which can be used to collect data and provide services to the users, 3) the algorithms which can be used to manage first person visual data for instance to perform localization, indexing, object detection, action recognition, human-object interaction and the prediction of future events.

Target Audience
First year PhD students, graduate students, researchers, practitioners.

Prerequisite Knowledge of Audience
Fundamentals of Computer Vision and Machine Learning (including Deep Learning).

Detailed Outline
The tutorial is divided into two parts and will cover the following topics:
Part I: History and motivation
• Agenda of the tutorial;
• Definitions, motivations, history and research trends of First Person (egocentric) Vision;
• Seminal works in First Person (Egocentric) Vision;
• Differences between Third Person and First Person Vision;
• First Person Vision datasets;
• Wearable devices to acquire/process first person visual data;
• Main research trends in First Person (Egocentric) Vision;
Part II: Fundamental tasks for first person vision systems:
• Localization;
• Hand/Object detection;
• Attention;
• Action/Activity recognition;
• Action anticipation;
• Industrial Applications.

A large number of 3D facial animation techniques have emerged with different goals, to make the representation of facial expressions more realistic, decrease computing time, the need for specialized equipment, etc. With new techniques come new definitions, concepts, and terms that correlate with methods that have existed for decades or even more recent ones. Parameterization, interpolation, blendshapes, motion-capture and others are concepts that often appear in the literature, in a generic way, as techniques, but which actually have different levels in information hierarchy. The first parte o this tutorial aims to clearly classify the different techniques and concepts of the 3D facial animation literature, locating them in each step of the 3D facial animation pipeline, through a parametric analysis.
The second part of the proposed tutorial will focus on Facepipe, an efficient and low cost alternative for capturing face motion information and applying it on 3D facial animation software. The proposed solution will be compared to existing ones regarding cost, features and utilization. At last, a hands-on step-by-step use of Facepipe will be performed to demonstrate the possibilities it brings to the 3D animation community.

Keywords
3D avatars, 3D facial animation, facial movement analysis.

Aims and Learning Objectives
The participants will learn about the most used alternatives for 3D facial animation and the main differences between them, in tandem with their advantages and disadvantages. Specifically, the participants will learn about: 1) A review regarding the 3D facial animation pipeline, 2) the devices that can be used to collect facial data and the existing solutions to perform facial capture and then apply it to 3D character animation, 3) a step-by-step guide on how to use Facepipe to perform facial capture and 3D facial animation.

Target Audience
PhD students, graduate students, researchers, designers, animators.

Prerequisite Knowledge of Audience
Fundamentals of 3D modeling (Blender will be used).

Detailed Outline
The proposed tutorial is divided into two parts and will cover the following topics:

Part I: A review regarding the 3D facial animation pipeline
• Review methodology
• The Facial Action Coding System (FACS)
• Intermediate Techniques
• Input Techniques
• Parametric Analysis of techniques

Part II: Facepipe and its utilization
• How Facepipe works
• Capturing data using Facepipe
• Using captured data on Blender
• Step-by-step guide

The main privilege of context-aware applications is to provide tailored services by analyzing the environmental context, such as location, time, weather condition, and seasons, and adapting their functionality according to the changing situations in context data without explicit user interaction. For example, mobile devices can obtain context information in various ways in order to provide more adaptable, flexible and user-friendly services. In case of a tourist app, a tourist would like to see relevant tourist attractions on a map together with distance information, depending on its current location. Human robots, require sensors to gather information about the conditions of the environment to allow the robot to make necessary decisions about its position or certain actions that the situation requires. As a consequence, context-aware applications can sense clues about the situational environment making applications more intelligent, adaptive, and personalized.
Visual Programming Languages (VPL) let users develop software programs by combining visual program elements, like sensor and actuator objects, loops or conditional statements rather than by specifying them textually. VPLs are considered to be innovative approaches to address the inherent complexity of developing programs, especially for beginners.
The tutorial will cover important technologies which can be used to build user-centered context-based applications, discussing challenges and open problems and will give conclusions and insights for research in the field. Several concrete scenarios will be analyzed and corresponding mobile applications will be developed during the tutorial by the participants using a cloud-based visual programming environment.

https://drive.google.com/drive/folders/1RnSbUnR9b_9f-V_FbPYFHiudUHq-v87c

Keywords
Context-aware Services; Sensors; Visual Programming; Mobile Applications; Location-based Services; Robot Applications; Humanoid Robots.

Aims and Learning Objectives
Participants will understand how modern technologies (e.g. NFC, BLE, Machine Learning) can provide important contextual information to develop user-centered context-based apps. Building on this, participants will learn how to develop context-based mobile applications using visual programming. For this purpose, several concrete scenarios will be analyzed and corresponding mobile applications will be developed during the tutorial. The participants will understand the main advantages of user-centered context-aware mobile applications and will be able to design user-centered context-based mobile applications by using visual programming.

Target Audience
Researchers, practitioners.

Prerequisite Knowledge of Audience
Fundamentals of Computer Science Concepts & Programming Language Concepts.

Detailed Outline
The tutorial is divided into three parts and will cover the following topics:

Part 1: User-centered context-aware applications 
- User-centered context-aware applications
- Categories of context information that are practically significant
- Technologies for context-aware applications (QR Codes, NFC, BLE, Machine Learning)
- Case Studies: Mobile Applications and Humanoid Robot Applications

Part 2: Visual Programming
- Visual programming vs. traditional programming
- Visual elements
- Cloud-based development environments: MIT App Inventor, Thunkable

Part 3: Case Study:
- Development of a user-centered context-based mobile application
- Development process using visual programming
- Analysis & design of a location-based mobile service by using a cloud-based development environment.

The tutorial will cover main technologies which can be used to build context-aware applications, discussing challenges and open problems and will give conclusions and insights for research in the field.

Grid-based layout algorithms have a wide range of applications in visual computing and computer graphics. These techniques are invaluable for tasks such as visualizing high-dimensional data, visually sorting images, and solving optimization problems. In this tutorial, we will focus on managing large amounts of textures or images that can overwhelm human perception. We begin with an overview of dimensionality reduction techniques and an examination of their inadequacy for sorting non-point data such as images or textures. The principles and concepts of various image sorting techniques are explained, including Self-Organizing Maps (SOM), Self-Sorting Maps (SSM), and the new innovative Linear Assignment Sorting (LAS). We also present neural networks for learning latent permutations. A major challenge in image sorting is the lack of appropriate metrics for evaluating sorting quality as perceived by humans. Based on extensive user testing, we present the novel Distance Preservation Quality metric, which shows a stronger correlation with user-perceived sorting quality compared to other metrics. Efficiency is another focus of this tutorial. This includes techniques such as integral image filtering and the use of linear assigments solvers for fast sorting. We also offer practical tips and tricks for achieving specific layout shapes and positioning specifications.
It's important to note that this tutorial goes beyond static image sorting to include dynamic image graph visualization that adapts to the ever-changing nature of image collections. Furthermore, we show how these techniques can be effectively applied to classical dimension reduction or optimization tasks, such as the Traveling Salesman Problem.

Keywords
Grid-based arrangements, Image sorting, Dimensionality reduction, Optimization techniques, Exploratory Image Search, Distance Preservation Quality Metric.

Aims and Learning Objectives
The participants will understand the basics and advantages of grid-based sorting. They will explore different sorting algorithms and draw insights from user testing to determine appropriate metrics for evaluating different sorting methods. Participants will also be introduced to the simple and efficient Linear Assignment Sorting technique, along with optimization strategies and practical programming examples. By the end of the workshop, attendees will have the skills and knowledge necessary to implement image sorting methods while taking into account specific layout requirements. In addition, the workshop will enable participants to adapt these algorithms to various optimization challenges.

Target Audience
First year PhD students, graduate students, researchers, practitioners.

Prerequisite Knowledge of Audience
Fundamentals of Computer Vision and Machine Learning.

Detailed Outline
- Welcome, gather participant information, and provide an overview of tutorial goals, schedule, and materials.
- Discuss the limitations of dimensionality reduction techniques (e.g., PCA, MDS, Isomap, LLE, t-SNE) for image organization.
- Introduce image sorting concepts, including SOM, SSM, LAS, neural networks for permutations.
- Examples of optimization and dimensionality reduction using the new algorithms.
- Highlight appropriate visual feature vectors for image sorting.
- Present insights into human perception and quality evaluation metrics.
- Explain optimization techniques for faster image sorting.
- Provide tips for dealing with layout constraints and fixed positioning during image sorting.
- Explain techniques for dynamic image graph visualization.
- Provide examples of industrial applications.
- Conclude with a final Q&A and recap of the key points of the tutorial.