Seminar: Computer Vision and Deep Learning for Autonomous Driving

PD Dr. Federico Tombari, Nikolas Brasch, Shun-Cheng Wu, Mahdi Saleh, Fabian Manhardt, Yanyan Li
Type: Seminar Module IN0014
SWS: 2+0
ECTS: 5 Credits
Location: Zoom meeting
Time: tba
Course Language: English

Announcements

• You can submit your motivation letter here.
• Preliminary Meeting: *July 14th, 11:00-11:30am, Zoom meeting PW: 479686 slides

Introduction

• The seminar includes a selection of the most recent and relevant papers in the field of computer vision and deep learning aimed at autonomous driving.
• Papers are selected to cover different aspects of the topic:
  • Semantic Segmentation
  • 3D Scene Understanding
  • Multi-modal sensor fusion
  • SLAM and 3D reconstruction
  • 6D object pose estimation / 3D object recognition
  • Joint depth and visual odometry prediction
  • Synthetic dataset creation
  • Generative models for sym-to-real
  • Semi-supervised large-scale dataset annotations
  • Trajectory planning and lane change prediction
• The full list of papers will be available a few weeks before the beginning of the semester.

Registration

• Registrations will be handled through the TUM Matching Platform.

Requirements

In this Master Seminar (formerly Hauptseminar), each student is asked to select one paper from a list. In order to successfully complete the seminar, participants have to fulfill these requirements:

• Presentation: The selected paper is presented to the other participants (20 minutes presentation 10 minutes questions). Use the CAMP templates for PowerPoint camp-tum-jhu-slides.zip, or Latex: CAMP-latex-template.
• Blog Post: Will be clarified in the first session.
• Attendance: Participants have to participate actively in all seminar sessions.

The students are required to attend each seminar presentation which will be held during this course. Each presentation is followed by a discussion and everyone is encouraged to actively participate. The blog post must include all references used and must be written completely in your own words. Copy and paste will not be tolerated. Both blog post and presentation have to be done in English.

List of Topics and Material

A list of papers will be announced here.

Nr	Authors	Title	Topic	Conference	Link	Tutor	Student
1	Yang et al.	3DSSD: Point-based 3D Single Stage Object Detector	Object detection	CVPR 2020	paper	Mahdi Saleh	Ebraheem
2	Najibi et al.	DOPS: Learning to Detect 3D Objects and Predict their 3D Shape	Object detection	CVPR 2020	paper	Mahdi Saleh	Fabian
3	Zhou et al.	Tracking Objects as Points	Object tracking	ECCV 2020	paper	Alexander Lehner	Tim
4	Bijelic et al.	Seeing Through Fog Without Seeing Fog: Deep Multimodal Sensor Fusion in Unseen Adverse Weather	Sensor Fusion	CVPR 2020	paper	Markus Herb	Gabriel
5	Zakharov et al	Autolabeling 3D Objects with Differentiable Rendering of SDF Shape Priors	Detection + Reconstruction	CVPR 2020	paper	Fabian Manhardt	Min Shan
6	Yang et al.	SurfelGAN?: Synthesizing Realistic Sensor Data for Autonomous Driving	Reconstruction + Completion	CVPR 2020	paper	Yanyan Li	Surendran
7	Guizilini et al.	Semantically-Guided Representation Learning for Self-Supervised Monocular Depth	Depth + Odometry	ICLR 2020	paper	Evin Pinar Örnek	Enis
8	Xue et al.	Learning Multi-view Camera Relocalization with Graph Neural Networks	Relocalization	CVPR 2020	paper	Shun-Cheng Wu	Frithjof
9	Zeng et al.	DSDNet: Deep Structured self-Driving Network	Motion planning	ECCV 2020	paper	Nikolas Brasch	Matthias
10	Cheng et al.	Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation	Panoptic Segmentation	CVPR 2020	paper	Markus Herb	Muhammed
11	Hu et al.	Probabilistic Future Prediction for Video Scene Understanding	Semantics/Depth/OpticalFlow	ECCV 2020	paper	Azade Farshad	Cecilia
12	Hugues et al.	KPConv: Flexible and Deformable Convolution for Point Clouds	Pointcloud Segmentation	ICCV 2019	paper	Stefano Gasperini	Pascal
13	Liu et al.	Morphing and Sampling Network for Dense Point Cloud Completion	Pointcloud Completion	AAAI 2020	paper	Yida Wang	Lukas
14	Parl et al.	Contrastive Learning for Unpaired Image-to-Image Translation	Domain Adaptation	ECCV 2020	paper	Artem Savkin	Erkam

Schedule

The schedule for the paper presentations will be announced here during the semester. Each student is required to attend all presentations.

Date	Time	Place	Speaker	Topic	Invited External Talk
Tue, July 14, 2020	11:00-11:30	Zoom	Federico	Preliminary meeting slides
Fri, Nov 6, 2020	16:00-17:00	Zoom	Federico	Introductory Session slides video
Fri, Nov 20, 2020	16:00-18:00	Zoom	Ebraheem, Fabian, Tim	Object Detection & Tracking	Peter Ondrúška Head Of Research at Lyft Level 5 Lyft
Fri, Nov 27, 2020	16:00-18:00	Zoom	Gabriel, Min Shan, Surendran	Sensor fusion, 3D Reconstruction
Fri, Dec 4, 2020	16:00-18:00	Zoom	Enis, Frithjof, Matthias	Motion estimation, Planning, Relocalization	Adrien Gaidon Senior Manager for Machine Learning Research Toyota Research Institute
Fri, Dec 11, 2020	16:00-18:00	Zoom	Muhammed, Cecilia	Segmentation	Thomas Sayre-McCord Advanced Autonomy Function Lead Lilium
Fri, Jan 15, 2020	16:00-18:00	Zoom	Lukas, Erkam	Completion, Domain Adaptation