You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Seminar/ Colloquium, which will be held on:
Friday, May 25th, 2018, 15:45-17:45h, at EWI-Lecture Hall Pi.
The programme features one MSc graduation project midterm presentations and a presentation of a research project by one of our PhD students.
Presenter 1: Yunchao Yin
Title: Annotation of cerebral angiography,
Abstract: Cerebral angiography is a medical imaging technique that used to visualize the vessels around the brain and provide quantitative measures for pathological changes such as arteriovenous malformations and aneurysms. However, it’s difficult for patients and young clinical stuff to tell the name of each vessel in the angiography. This project plans to create an automatic cerebral vessel name annotation tool based on deep learning. Semantic segmentation and topological skeleton extraction are both possible to realize automatic vessel names annotation, but the ground truth for semantic segmentation is relatively hard to get and pixel-wise perfect is not required for this task, so skeleton extraction is chosen. The project could be divided into two part: 1) angiography acquirement angle classification; 2) Vessel bifurcation detection. Different cerebral vascular model is used for angiography annotation of different projection angles, and the second part of the project predicts vessel names.
Presenter 2: Chaoran Fan
Title: Fast and accurate CNN-based brushing in scatterplots
Abstract: Brushing plays a central role in most modern visual analytics solutions and effective and efficient techniques for data selection are key to establishing a successful human-computer dialogue. With this paper, we address the need for brushing techniques that are both fast, enabling a fluid interaction in visual data exploration and analysis, and also accurate, i.e., enabling the user to effectively select specific data subsets, even when their geometric delimination is non-trivial. We present a new solution for a near-perfect sketch-based brushing technique, where we exploit a convolutional neural network (CNN) for estimating the intended data selection from a fast and simple click-and-drag interaction and from the data distribution in the visualization. Our key contributions include a drastically reduced error rate—now below 3%, i.e., less than half of the so far best accuracy— and an extension to a larger variety of selected data subsets, going beyond previous limitations due to linear estimation models.
You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Seminar/ Colloquium, which will be held on: Friday, April 13th, 2018, 15:45-16:45h, at EWI-Lecture Hall Pi. The programme features a guest talk.
Presenter: Dr. R.C. Lindenbergh
TU Delft, Dept. of Geoscience & Remote Sensing
Title: Robust feature extraction and change detection in large spatial point clouds
Abstract: Laser scanning is an efficient method to sample our urban and natural environment. LIDAR systems on tripods, drones, cars, drones and planes are able to collect billions of 3D points in a few hours. What remains challenging is to extract valid metric information from these points clouds in a time that matches the acquisition time. In the presentation first issues with these point clouds will be discussed, followed by example methods to extract metric information on e.g. trees, traffic signs or tunnels in a robust and computationally efficient way.
On 15th of March, 2018, Jan-Willem van Velzen successfully defended his master thesis titled “Texture-based Rendering of Vector-based Shapes”. In his work he presents an approach to store vector-based shapes into discrete raster-based textures, optimized for parallel rendering. He also discusses the implementation of both the serialization and deserialization steps, as well as rendering regular vector shapes with a comparable quality.
We congratulate Jan-Williem with his work and defense and wish him all the best for the future!
You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Seminar/ Colloquium, which will be held on: Friday, March 16th, 2018, 15:45-16:45h, at EWI-Lecture Hall Pi. The programme features a MSc graduation project midterm presentations.
Presenter: Niels van der Veen
Title: A paint-based approach for optimal lighting design in real scenes
Abstract: Light design is a computational expensive task, commonly done using Computer-aided design (CAD) software in a virtual scene. The designer places and tunes the virtual light sources and, yet the virtual environment is ideal for physically correct light tracing, the costly simulation might not provide the desired impression in the real scene. Moreover, the chosen light placement is not necessarily optimized. In this work we capture light behavior from real scenes as well as 3D scene properties and use this information to recreate different lighting designs. The results approximates physical correctness while visualizing the illumination on the real scene in a more time-efficient way. To make the design process more intuitive, the user paints the desired light properties instead of placing the light sources. The system attempts to find the optimal positions and parameters of the light sources in the scene to reflect the current design. Constraints such as number of light sources, emission profiles and external light information can also be specified.
The meeting on March 2nd is scheduled from 15:45-16:45 in EWI lecture hall Pi. The programme features a MSc graduation project midterm presentations.
Presenter: Bastiaan Grisèl
Title: Interactive analysis of 3D embeddings in Virtual Reality
Abstract: Developments in algorithm design and computing power have enabled dimensionality reduction algorithms to run in real-time. While embedding data in three dimensions can lead to a more faithful representation of a dataset when compared to embedding in two dimensions, embeddings are often visualised in 2D (on computer screens) with limited possibilities for interaction. Virtual Reality (VR) makes it possible to visualise embeddings in stereo 3D and interact with them using VR motion controllers, enabling the user to interact with the algorithm as it evolves and develop an understanding of both the algorithm and the dataset. This research uses a well-known dimensionality reduction algorithm called t-SNE and aims to provide the user with virtual tools to work with the algorithm to generate more powerful embeddings and better understand high-dimensional datasets.
You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Seminar/ Colloquium, which will be held on:
Friday, February 2nd, 2017, 13:45-15:45h, at EWI-Lecture Hall K.
The program features the following two presentations:
Presenter: Yin Yunchao
Title: Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning
Abstract: Remarkable progress has been made in image recognition, primarily due to the availability of large-scale annotated datasets and deep convolutional neural networks (CNNs). CNNs enable learning data-driven, highly representative, hierarchical image features from sufficient training data. However, obtaining datasets as comprehensively annotated as ImageNet in the medical imaging domain remains a challenge. There are currently three major techniques that successfully employ CNNs to medical image classification: training the CNN from scratch, using off-the-shelf pre-trained CNN features, and conducting unsupervised CNN pre-training with supervised fine-tuning. Another effective method is transfer learning, i.e., fine-tuning CNN models pre-trained from natural image dataset to medical image tasks. In this paper, we exploit three important, but previously understudied factors of employing deep convolutional neural networks to computer-aided detection problems. We first explore and evaluate different CNN architectures. The studied models contain 5 thousand to 160 million parameters, and vary in numbers of layers. We then evaluate the influence of dataset scale and spatial image context on performance. Finally, we examine when and why transfer learning from pre-trained ImageNet (via fine-tuning) can be useful. We study two specific computer-aided detection (CADe) problems, namely thoraco abdominal lymph node (LN) detection and interstitial lung disease (ILD) classification. We achieve the state-of-the-art performance on the mediastinal LN detection, and report the first five-fold cross-validation classification results on predicting axial CT slices with ILD categories. Our extensive empirical evaluation,CNN model analysis and valuable insights can be extended to the design of high performance CAD systems for other medical imaging tasks.
Presenter: Dirk Schut
Title: Automatic Initialization for US CT Registration During Liver Tumor Ablations
Abstract: Ablation is a medical procedure where a needle has to be inserted into a tumor. To guide the needle Ultrasound(US) imaging is used which shows the needle position in real time. However small tumors are often not visible on US images. Therefore a CT scan is acquired before the intervention. To make the information from the CT scan available during the intervention, the scans have to be aligned in such a way that the same parts of the body are visible in the same positions in both scans. To find the relative orientation between the scans, this technique tries to detect and match blood vessels and the liver surface while also taking into account that the US scanner should be on the skin of the patient. The technique is being developed specifically to find solutions over a wide search space, and to be robust to segmentation errors.
You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Colloquium, which will be held on Tuesday March 15, 2016, 14:00-15:00, in the Vassiliadis zaal (HB10.230).
- Dr. Amir Vaxman, Utrecht University
- Title – Directional Field Synthesis, Design, and Processing
- Abstract – Directional fields on discrete surfaces and in volumes are key components of geometry processing. Many applications make use of such fields, among which are remeshing, surface parametrization (texture mapping), texture synthesis, fluid simulation, and many more. I will present the challenges and the limitations in the design and the analysis of such fields, and focus on novel ways to compute them.
You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Colloquium, which will be held on Thursday July 16, 2015, 15.45-17.20, in the Vassiliadis zaal (HB10.230). The program this month features the following speakers:
- Panchamy Krishnan
- Title – Supervised learning for measuring hip joint distance in digital X-ray images
- Abstract – Osteoarthritis is a degenerative joint disease which is hard to measure objectively as the diagnosis differs from surgeon to surgeon. Hip X-rays are used to diagnose this disease by measuring several characteristic features of the image, mainly the joint distance between the femoral head and acetabular cup. Joint distance reduction is a clear sign of OA (it is a symptom of cartilage disappearance). This metric involves segmentation of the femur and acetabulum in x-rays, a challenging task because of contrast variations as well as external factors like anatomical and pose-variation. A multi-scale approach based on Machine Learning is proposed here providing a robust segmentation by detecting landmarks along the edges of the bone. The labelled landmarks are then used to determine the joint distance in several locations along the hip joint. The learning is supervised using manually annotated landmarks. The impact of landmark selection is studied in detail in this paper. The landmark detection is further refined using Active Shape Model (ASM) and gradient profiling.
- Bas Dado
- Title – Compression of voxelized scenes using colored DAGs
- Abstract – Voxels are great alternative for triangle-meshes. No texture and bump-mapping techniques are necessary to capture fine details, and rendering by means of raycasting is relatively cheap. However, a big dowside of voxels is the large memory footprint, even when a Sparse Voxel Octree (SVO) is used. Directional Acyclic Graphs can be used to reduce this problem but don’t allow for storing materials. We present a technique, based on the DAG compression, to efficiently store the voxel data. The presented technique is smaller than a standard SVO for all tested scenes in high resolution.
- Kevin Allain
- Title – Enhancing document triage process using multi layer visualization tool
- Abstract The European Patent Office is an international organisation that is charged with granting patents for applications sent to their offices. To do so, the examiners have to verify if the claims of the application are indeed new or not, and thus perform a state of the art review. The tools currently used at the EPO return a list of patents that might match the keywords entered by the examiner, but it provides a very poor visualization of the results and document triage is a very cumbersome process. We work on a visualization of all the results from a search query to increase the document triage process, by showing data of all the results at once, offering several levels of details for the examiner to determine the patents that might not be relevant.
As a reminder, all CGV MSc students (both seminar and thesis project) are expected to attend our monthly colloquium series. All other interested colleagues are more than welcome, so feel free to disseminate this announcement.
You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Colloquium, which will be held on Thursday June 18, 2015, 15.45-17.20, in the Vassiliadis zaal (HB10.230).
The program this month features the following speakers:
- Hongjie Li
- Title – Automated Expansion of Statistical Shape Model Library for femur
- Abstract – Currently Clinical Graphics has a Statistical Shape Model (SSM) trained from 275 3D meshes. The SSM can be used to segment MRI data of femur and hip. They also have 2000 partial femur meshes and want to incorporate the partial femur to the SSM training set, so that more reasonable variance can be introduced. My task it set one criteria to decide whether accept and reject the partial femur. Then incorporate the partial bone to the training set, and evaluate whether the new SSM improve the fit accuracy.
- Fieke Taal
- Title – Procedural Generation of Traffic Signs
- Abstract – Designing realistic 3D content in virtual simulations is a hot topic and a challenge, especially in the context of virtual traffic situations considering several users, objects and traffic rules. Manually generating these virtual environments is not efficient, so procedural methods are needed to create these complex traffic models. The goal of this project is to come up with a procedural method to generate automatically the right traffic signs in a road network.
- Bas in het Veld
- Title – Procedural generation of populations for storytelling
- Abstract – Procedural world generation is often limited to creating worlds devoid of people and any background. Because of this, creating a vibrant, living world is still a problem that requires a skilled designer. We present a method that generates a socially connected population in any virtual terrain, using a mixed-initiative simulation of settlements that adapt to the world and to a designer’s input. Using this simulation, we develop a number of sample worlds that convey the expressive potential of the approach.
As a reminder, all CGV MSc students (both seminar and thesis project) are expected to attend our monthly colloquium series. The next Colloquium is planned on Thursday the 16th of July. All other interested colleagues are more than welcome, so feel free to disseminate this announcement.
You are cordially invited to attend our next Computer Graphics and Visualization (CGV) Colloquium, which will be held on Thursday May 21, 2015, 15.45-17.20, in the Vassiliadis zaal (HB10.230).
The program this month features the following speakers:
- Kai Lawonn
- Title – Illustrative Visualization of Medical Data Sets
- Abstract – The aim of an illustrative visualization method is to provide a simplified representation of a complex scene or object. Concave and convex regions are emphasized and the surface complexity is reduced by omitting unnecessary information. This abstraction is often preferred over fully illuminated scenes in a multitude of applications. This talk presents an overview of the state-of-the-art for feature lines. Additionally, an evaluation is shown to assess the quality of feature lines on anatomical data sets. Based on the evaluation, two conclusions in the field for medical applications were derived. From this point, this talk presents two solutions in the field of illustrative visualization for medical data sets. A novel line drawing technique will be presented to illustrate surfaces. According to different requirements, this technique will be extended for molecular surfaces. In the field of vessel visualization with embedded blood flow, an adaptive visualization method will be introduced. This technique will also be extended to animated blood flow. Finally, this talk shows different illustrative visualization concepts, which can be applied in various fields for depicting surface information.
- Renata Raidou
- Title – Test Talk EuroVis2015: Visual Analytics for the Exploration of Tumor Tissue Characteristics
- Abstract – Tumors are heterogeneous tissues consisting of multiple regions with distinct characteristics. Characterization of these intra-tumor regions, using medical imaging data, can improve patient diagnosis and enable a better targeted treatment. However, the high dimensionality and complexity of the imaging-derived feature space of tumor tissue characteristics is prohibiting for easy exploration and analysis. We propose a visual tool for: (1) easy exploration and visual analysis of the feature space of imaging-derived tissue characteristics and (2) knowledge discovery and hypothesis generation and confirmation, with respect to reference data used in clinical research. We employ, as central view, a 2D embedding of the imaging-derived features. Multiple linked interactive views provide functionality for the exploration and analysis of the local structure of the feature space, enabling linking to patient anatomy and clinical reference data.
- Kristinn Rúnarsson
- Title – Animated Photomosaics
- Abstract – A photomosaic is an image that has been divided into a grid of tiles, each of which is then replaced with another image that matches the tile section best. Animated photomosaics extend the previous photomosaic techniques to support video clips as input. The goal of the project is to come up with image descriptor suitable to generate photomosaics, and an application to efficiently create visually pleasing and coherent animated photomosaics.
As a reminder, all CGV MSc students (both seminar and thesis project) are expected to attend our monthly colloquium series. The next Colloquium is planned on Thursday 18 June. All other interested colleagues are more than welcome, so feel free to disseminate this announcement.