Rules and requirements
The goal of the challenge is to foster research on machine learning for 3D audio. All participants should adhere to the following rules to be eligible for the challenge:
- All participants must submit the obtained results for at least one of the 2 tasks and one track (audio-only and audio-visual), but for both sub-tracks (1 and 2 mics). The results should be accompanied by a paper describing the proposed method.
- Each individual participant cannot be included in multiple participating teams. Therefore, a participant is allowed to submit only one set of results.
- There are no restrictions on the proposed methodologies. However, in case of a tie, the Challenge Committee will take into account the novelty and originality of the proposed approach. Also, a method that can be used for both the 1-mic and 2-mic configurations will be positively evaluated.
- Participants are not restricted to use the L3DAS23 dataset only. It is in fact allowed to augment this dataset and/or to integrate additional publicly available data to train/pre-train the models.
- The Challenge can have up to 5 papers from the top ranked teams. The format should be consistent with a 2-page ICASSP paper, and should be submitted before the camera-ready deadline.
- Accepted papers will be presented at a Signal Processing Grand Challenge special session of the IEEE ICASSP 2023 conference.
- Intellectual property (IP) is not transferred to the challenge organizers, i.e., if code is shared/submitted, the participants remain the owners of their code.
- Each participant must provide the number of hours of model training and an estimate of carbon footprint, also calculated on platforms such as ML C02 Impact. While not assessable for final ranking purposes, we encourage participants to optimize the use of their computational resources and consider the impact of training sessions on the environment.