Goal
The doctoral researchers of all six Helmholtz Information & Data Science schools come together to work with drone images of thermal leakages in construction.
HIDA will provide a baseline computer vision model and their task will be to implement a semantic segmentation pipeline with the aim to improve as much as possible the precision of the 50% intersection-over-union of the thermal bridge bounding boxes.
Registration
Please first register via the Indico system (first-come-first-serve per HIDS School, if the hackathon is full, you can still register and will be put on the waiting list).
You will then have to register with the HAICORE compute resources@KIT by self sign-on explained here: https://www.nhr.kit.edu/userdocs/haicore/registration/
Please make sure to use the e-mail address from you Helmholtz center for registration. Should you not be directly affiliated with any center your authorization will take longer and you will be separately notified via mail.
We want to make sure that HAICORE registration and system setup are finalized before the event, so you will have to run a short test script on HAICORE upon first login. You then need to provide the personalized registration token you get from the script in the Indico system to finalize your registration.
Deadline for uploading the token is 1 week prior to the event (19th of February).
IMPORTANT:
If you have not uploaded your personalized HoReKa token, your registration will be incomplete and your spot will be given to someone on the waiting list.
Requirements
You should:
- be an advanced AI-researcher with experience in model development, training and optimization
- be familiar working either in Jupyter Hub or on a Linux shell
- optional: have some experience in HPC job submission via batch-systems (SLURM, LSF, etc.)
Agenda
Tuesday (Feb 27th), 09.30 am - Kick-off session with introductory talks
Thursday (Feb 29th), 4.00 pm - Wrap-up and winner announcement
The detailed agenda can be found here.
IMPORTANT:
Participation in the initial session is mandatory to all participants, as it will give you introduction to the challenges and datasets, and you will be assigned to your team. If you do not show up for this session, we cannot assign you to a team, and your participation is hence forfeit.
Evaluation
The winner of the challenge will be the team scoring the best average intersection-over-union value for all test images. Due to fairness, we will employ a strict non-cheating policy: team attempting to falsify the final scoring metric will be disqualified without an option to appeal.
