Semantic segmentation plays a pivotal role in enabling autonomous vehicles to understand their surroundings, but adverse weather conditions pose significant challenges, often leading to hazardous situations. The central challenge addressed by the competition is optimizing the Safe mIoU metric for semantic s egmentation in adverse weather conditions to ensure the safety of autonomous driving systems.
This competition is aimed at assessing the performance of semantic segmentation models in adverse weather conditions for autonomous driving. Participants are provided with the IDD-AW dataset, comprising 5000 pairs of high-quality images with pixel-level annotations, captured under adverse weather conditions such as rain, fog, low light, and snow. Emphasizing the correct identification of safety-critical elements and penalizing unsafe false predictions (False). The Participants aim to develop robust models that accurately segment driving scenes even in adverse weather conditions, focusing on prioritizing safety-related mispredictions.
Name | mIoU % | SmIoU (tp) % | SmIoU % |
---|---|---|---|
KweenCoders | 68.32 | 64.73 | 53.08 |
SixthSenseSegmentation | 68.54 | 63.05 | 51.16 |
SemSeggers | 67.58 | 62.15 | 50.98 |
Anidh & Krishna | 66.49 | 61.85 | 50.10 |
IDDAW Paper | 64.70 | 60.56 | 51.32 |
Sanket | 64.82 | 60.22 | 48.86 |
xmba15 | 32.18 | 17.91 | 2.10 |
On completion of the task, we aim to designate winners based on the evaluation process of the proposed system. Winning teams that successfully complete the task will receive prizes as follows:
Standing | Reward |
---|---|
1st Place | 1000 USD |
2nd Place | 500 USD |
3rd Place | 300 USD |