Project Materials

Booklet - Presentations - Posters - Video

Booklet

The poster booklet contains general KI Data Tooling information, illustrates the project's approaches, and presents research results.

>>> Download Booklet: KI Data Tooling - The Data Kit for Automotive AI <<<

Presentations for Download

Presentations of the remarkable KI Data Tooling Final Event on project improvements, challenges on the KI Data Tooling journey and opportunities ahead; interesting Keynotes and notable Deep Dives.

An audio file can be provided, please contact us: ki-dt-projektmanagement(at)reach.eict.de

Keynote: Next Level Data KIT | Stefan Voget, Continental
KI Data Tooling: Project Overview, Challenges and Stories to tell | Armin Koehler, Bosch
Keynote: Standardisation of safe data-driven KI Development & Tooling | Simon Burton, Fraunhofer IKS
KI Data Tooling - Introducing KI-DT Data Kit Framework for building resilient automotive AI | Thomas Stauner, BMW; Armin Koehler, Bosch; Evren Ermis, Continental; Marcel Matz, Valeo; Thomas Stone, BMW; Maarten Bieshaar, Bosch
Keynote: From ChatGPT to GAIA-1: On Generative Sequence Models in Speech, Language, and Vision | Further information on RessearchGate (external link) | Tim Fingscheidt, TU Braunschweig
KI Data Tooling - Data centric AI Developer Journey in the KI-DT Framework | Thomas Stone, BMW, KI-DT Experts
Deep Dive #1 Training with Synthetic Data – Mixed Training | Maarten Bieshaar, Maximilian Menke, Bosch
Deep Dive #2 Real Data & Labeling | Andrea Kraus, Valeo
Deep Dive #3 Synthetic Data Production & Validation | Armin Köhler, Bosch; Christian Holzapfel, Dominik Salles, FKFS; Thorsten Püschl, dSPACE; Günther Hasna, Ansys; Ludwig Friedmann, BMW; Tobias Denk, BIT-Technology Solutions; Nikolas Hemion, dSPACE
Deep Dive #4 Corner Case | Florian Heidecker, Uni Kassel; Isabelle Groß, FKA (ZF); Tianming Qiu, Fortiss (BMW); Jasmin Breitenstein, TUBS; Kamil Kowol, BUW
KI Data Tooling: Big Picture for Future Development | Hans-Jörg Vögel, BMW

Posters for Download

1 Scenarios & Synthetic Data

Creation of Digital Twins of Urban Traffic Spaces for Sensor Simulation and Synthetic Data Generation
3D Asset Creation for Synthetic Data Production
Material Parameters in Radar Simulation
Validation Tools for 3D Models and Materials
Converting Real-World Measurements into Scenario-Based Simulations
Camera Challenges and Evaluation of Camera Specific Properties
Radar Sensor Simulation
AI Use Cases with Radar Data: An Overview
The Stuttgart Driving Simulator for Driver-in-the-Loop Simulation Tests
CARLA-Wildlife: A Synthetic Video Data Set for Tracking and Retrieval of Out of Distribution Objects

2 Acquisition, Curation & Refinement of Real Data

Context Information for Corner Case Detection in Highly Automated Driving
Multi Trajectory and Context Recording
Benefits of the “AIM Research Intersection” for KI Data Tooling
Context Generation from Images and Structured Data
Latent Diffusion Face Anonymization
About the Ambiguity of Data Augmentation for 3D Object Detection
Synthesizing and Adapting Virtual Humans in 3D Environments
From Sensors to Scene Understanding via Auto-Labeling
Creating a Multimodal Sensor Data Set

3 Data Storage, Analysis & Discoverability

Focus on the Challenges: Analysis of a User-Friendly Data Search Approach with CLIP in the Automotive Domain
Augmentation of Images with Generative Networks
Scalable Data Set Distillation
A-Eye: Driving with the Eyes of AI for Corner Case Generation
Exploring the Unknown: Active Learning via Neural Network Uncertainty Modeling
Adaptive Bitrate Quantization Scheme Without Codebook for Learned Image Compression
Corner Case Identification Using Cameras and GPS
Self-Annotated Discovery of 3D Instances in Outdoor Scenes

4 ML Function Development

Sensor Equivariance
Generation and Detection of Corner Cases
Criteria for Uncertainty-Based Corner Cases Detection in Instance Segmentation
Relevance Estimation of Corner Cases for Semantic Segmentation
Domain Adaptation with cDCGAN for Semantic Segmentation
Unsupervised Domain Adaptation for Object Detection Using Adversarial Style Transfer and Semi-Supervised Learning
Joint Prediction of Amodal and Visible Semantic Segmentation for Automated Driving
Towards an End-to-End Amodal Video Instance Segmentation Challenge
Mixed Training - Identification and Filling of Data Gaps
Mixed Data Set Training 
Real-Synthetic Data Mix Training
Probabilistic Trajectory Forecast of VRUs

Video

>>> Download Video: KI Data Tooling - The Data Kit for Automotive AI (720p/mp4) <<<