Instructors: Prof. Dr. Nassir Navab, Anees Kazi, Mahsa Ghorbani, Matthias Keicher, Tamara Mueller, Kamilia Mullakaeva, Tobias Czempiel, Ege Ozsoy, Felix Holm
Announcements
- 26.04.2022 - Seminar starting date
- 06.04.2022 - List of papers will be announced
- 23.03.2022 - Deadline for deregistration
- 08.03.2022 - Notification for acceptance to the seminar
- 03.02.2022 - Premilinary meeting
- 03.03.2022 - Contact information. If you have any questions about the seminar, feel free to contact Anees Kazi (anees.kazi@tum.de)
- 10.02.2022 - Registrations are open through the TUM Matching Platform. Additionally, let us know your interest through our application form,
- The maximum number of participants: 20.
- Join the lecture virtually at:
Zoom Meeting
https://tum-conf.zoom.us/j/69782153188Meeting ID: 697 8215 3188
Passcode: 095144
Introduction
Graph Deep Learning is a new exotic branch in many fields like computer Vision and Medical Imaging. Many real-world medical and non-medical datasets can be represented in the form of graphs, providing a powerful source of information for machine learning models. This graph-based data, combined with the success of the convolutional neural networks, has motivated to translate the key ingredients of deep learning models into the graph domain. Many communities such as healthcare, social media, and computer vision are moving towards analyzing the data using Graph Convolutions. This seminar provides a space for discussion of the recent scientific publications on GCN with a focus on their medical applications and others.
Registration
- Interested students should attend the introductory meeting on 03.02.2022 to enlist in the course. Details are available at TUM online.
- Students can only register from 10.02. bis 15.02.2022 through TUM Matching Platform themselves if the maximum number of participants hasn't been reached (please pay attention to the Deadlines).
- A maximum number of participants: 20.
Requirements and Course Structure
For this course, previous background in Machine/Deep Learning is required.
During the course, students will select and present a scientific publication from the given list. Additionally, students will also contribute to the course blog post with their own summary of the selected article. Requirements are described in the image below:
List of Topics
Topic | Number | Title | Journal/Conference | Tutor | Student | Link |
---|---|---|---|---|---|---|
Architecture Design | 1 | Improving Graph Neural Networks with Simple Architecture Design | Preprint | Franz Rieger | Berfin Elif Erdoğan | https://arxiv.org/pdf/2105.07634v1.pdf |
Road Graphs and Graph-Tensor Encodings | 2 | Sat2Graph: Road Graph Extraction through Graph-Tensor Encoding | ECCV 2020 | Johannes Paetzold | Marcos Balle Sanchez | https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123690052.pdf |
Brain Mapping | 3 | 2D histology meets 3D topology: Cytoarchitectonic brain mapping with Graph Neural Networks | MICCAI 2021 | Alina Dima | Merve Elif Demirtas | https://arxiv.org/pdf/2103.05259.pdf |
Registration | 4 | GraphRegNet: Deep Graph Regularisation Networks on Sparse Keypoints for Dense Registration of 3D Lung CTs | IEEE Transactions on Medical Imaging 2021 | Mohammad Farid Azampour | Ha Young Kim | https://ieeexplore.ieee.org/abstract/document/9406964 |
Registration | 5 | IGCN: Image-to-graph Convolutional Network for 2D/3D Deformable Registration | Preprint | Mohammad Farid Azampour | Ata Jadid Ahari | https://arxiv.org/pdf/2111.00484.pdf |
Scene Graphs | 6 | Spatial-Temporal Transformer for Dynamic Scene Graph Generation | ICCV 2021 | Ege Özsoy | Caghan Koksal | https://arxiv.org/pdf/2107.12309.pdf |
Scene Graphs | 7 | Target Adaptive Context Aggregation for Video Scene Graph Generation | ICCV 2021 | Ege Özsoy | Haichuan Li | https://arxiv.org/pdf/2108.08121.pdf |
Surgical Scene Graphs | 8 | Global-Reasoned Multi-Task Learning Model for Surgical Scene Understanding | ROL 2022 | Felix Holm | Yuhan Li | https://arxiv.org/pdf/2201.11957.pdf |
Surgical Scene Graphs | 9 | Learning and Reasoning with the Graph Structure Representation in Robotic Surgery | MICCAI 2020 | Felix Holm | Chengzhi Shen | https://arxiv.org/pdf/2007.03357.pdf |
Scene Graph Generation | 10 | RelTR: Relation Transformer for Scene Graph Generation | Preprint | Matthias Keicher | Priyank Upadhya | https://arxiv.org/abs/2201.11460 |
Medical Report Generation, Few-shot Learning | 11 | MetaDT: Meta Decision Tree for Interpretable Few-Shot Learning | Preprint | Matthias Keicher | Chaehyeon Sim | https://arxiv.org/abs/2203.01482 |
Anomaly detection, chest x-ray application | 12 | Chest Radiograph Disentanglement for COVID-19 Outcome Prediction | MICCAI 2021 | Kamilia Mullakaeva | Diyorbek Rustamov | https://arxiv.org/pdf/2105.09937.pdf |
GNN Architecture | 13 | Neural Trees for Learning on Graphs | NeurIPS 2021 | Evin Pınar Örnek | Niklas Vest | https://openreview.net/pdf?id=UwSwML5iJkp |
Self-supervision | 14 | Context Matters: Graph-based Self-supervised Representation Learning for Medical Images | AAAI 2021 | Mahsa Ghorbani | Xinyue Zhang | https://arxiv.org/pdf/2012.06457.pdf |
Anomaly detection using GCN | 15 | Airway Anomaly Detection by Graph Neural Network | MICCAI 2021 | Kamilia Mullakaeva | https://miccai2021.org/openaccess/paperlinks/2021/09/01/046-Paper0604.html | |
Explainability of GNNs | 16 | On Explainability of Graph Neural Networks via Subgraph Explorations | ICML 2021 | Tamara Mueller | Malika Sanhinova | http://proceedings.mlr.press/v139/yuan21c/yuan21c.pdf |
Scene Graphs | 17 | Decoupling Object Detection from Human-Object Interaction Recognition | Preprint | Tobias Czempiel | Xinyu Chen | https://arxiv.org/pdf/2112.09828.pdf |
Dynamic Scene Graphs | 18 | Exploiting Long-Term Dependencies for Generating Dynamic Scene Graphs | Preprint | Tobias Czempiel | https://arxiv.org/pdf/2112.09828.pdf | |
Pooling | 19 | Hierarchical Graph Representation Learning with Differentiable Pooling | Advances in neural information processing systems 2018 | Anees Kazi | Daniel Stoll | https://arxiv.org/abs/1806.08804 |
Uncertainty | 20 | Edge-variational Graph Convolutional Networks for Uncertainty-aware Disease Prediction | MICCAI 2020 | Anees Kazi | Rafael Cabral Muchacho | https://arxiv.org/abs/2009.02759 |
Schedule
Sessions will take place Tuesdays from 12:00 to 14:00 hrs. A group of Two to Three students will present their papers during each session.
Sessions will start on 26.04 with introductory lectures by the organizers. Room is 03.09.012.
Date | Time | Topic | Additional Info/ slides |
---|---|---|---|
26.04.22 | 12:00 - 14:00 | Lecture: Intro to GDLMA | |
03.05.22 | 12:00 - 14:00 | Improving Graph Neural Networks with Simple Architecture Design | |
Sat2Graph: Road Graph Extraction through Graph-Tensor Encoding | |||
Hierarchical Graph Representation Learning with Differentiable Pooling | |||
10.05.22 | 12:00 - 14:00 | On Explainability of Graph Neural Networks via Subgraph Explorations | |
MetaDT: Meta Decision Tree for Interpretable Few-Shot Learning | |||
Edge-variational Graph Convolutional Networks for Uncertainty-aware Disease Prediction | |||
17.05.22 | 12:00 - 14:00 | Talk 1: Dr. Seyed Ahmad Ahmadi (NVIDIA) | |
24.05.22 | 12:00 - 14:00 | Chest Radiograph Disentanglement for COVID-19 Outcome Prediction | |
Context Matters: Graph-based Self-supervised Representation Learning for Medical Images | |||
Airway Anomaly Detection by Graph Neural Network | |||
31.05.22 | 12:00 - 14:00 | 2D histology meets 3D topology: Cytoarchitectonic brain mapping with Graph Neural Networks | |
GraphRegNet: Deep Graph Regularisation Networks on Sparse Keypoints for Dense Registration of 3D Lung CTs | |||
IGCN: Image-to-graph Convolutional Network for 2D/3D Deformable Registration | |||
14.06.22 | 12:00 - 14:00 | Talk 2 : Alaa Bessadok (Helmholtz) | |
21.06.22 | 12:00 - 14:00 | Lecture: Intro to Scene Graphs | |
28.06.22 | 12:00 - 14:00 | Neural Trees for Learning on Graphs | |
RelTR: Relation Transformer for Scene Graph Generation | |||
Decoupling Object Detection from Human-Object Interaction Recognition | |||
5.07.22 | 12:00 - 14:00 | Spatial-Temporal Transformer for Dynamic Scene Graph Generation | |
Target Adaptive Context Aggregation for Video Scene Graph Generation | |||
Sat2Graph: Road Graph Extraction through Graph-Tensor Encoding | |||
12.07.22 | 12:00 - 14:00 | Global-Reasoned Multi-Task Learning Model for Surgical Scene Understanding | |
Learning and Reasoning with the Graph Structure Representation in Robotic Surgery | |||
19.07.22 | 12:00 - 14:00 | Talk 3 : Azade Farshad (CAMP) | |
26.07.22 | 12:00 - 14:00 | Paper session - Backup | |
Resources and Material
An introductory lecture to GCNs by Prof. Xavier Bresson on Alfredo Canziani Youtube channel:
If you find more interesting material you would like to share in this section, feel free to contact us.