Introduction:

The Electrooptical-Infrared (EO-IR) object detection challenge is dedicated to advanced object detection in thermal imaging, a critical modality for scenarios where visible-light sensors fail in conditions such as low-light, night time and adverse weather. Thermal imaging provides unique advantages by capturing long-wavelength infrared (LWIR) data, making it an essential tool for robust object detection in challenging environments.

Dataset Novelty:

Unlike other dominating datasets collected using drones for object detection, IndraEye offers multiple modalities and various slant ranges making it the only ideal dataset for task such as traffic management, surveillance, etc. This is the first of its kind in Asian sub-continent annotated for object detection and semantic segmentation. The dataset comprises of 3031 IR images and 2083 EO images. These frames include diverse scenes and object types in public spaces, providing a comprehensive benchmark for thermal object detection.

The Challenge:

Participants are invited to develop innovative algorithms that leverage both modalities (EO and IR) to improve object detection on both domains. The challenge emphasizes single-modality tracking in the thermal domain and encourages approaches that address the unique characteristics of thermal imagery, such as noise, resolution, and contrast. Training EO and IR can be done together or using approaches such as knowledge distillation, domain adaptation, etc,.

Submission will take place on the Kaggle platform: https://www.kaggle.com/competitions/etaav-eo-ir

Rules:

• Team size: Maximum 5 members per team.

• Pretrained Models: Use of pretrained models is permitted (Must be specified).

• Submission Limit: 5 submissions per day.

• Solutions using single stage approach i.e, YOLOv8 is recommended but the participants can also propose novel algorithms and will be evaluated based on parameters-to-performance ratio for a fair comparison.

• Models must not be trained on images from the validation set.

Disqualifications:

• Before the result announcement, we will conduct a final evaluation on test dataset to ensure the integrity of the competition. This helps us verify submissions and maintain a fair and competitive environment.

• Any signs of rule violations or inconsistencies may result in disqualification.

Citation:

If you aim to publish your work using this dataset, kindly consider citing the paper using:

@misc{d2025sagasemanticawaregraycolor,

     title={SAGA: Semantic-Aware Gray color Augmentation for Visible-to-Thermal Domain Adaptation across Multi-View Drone and Ground-Based Vision Systems},

     author={Manjunath D and Aniruddh Sikdar and Prajwal Gurunath and Sumanth Udupa and Suresh Sundaram},

     year={2025},

     eprint={2504.15728},

     archivePrefix={arXiv},

     primaryClass={cs.CV},

     url={https://arxiv.org/abs/2504.15728},

}

Outcome:

• Paper submission: The selected teams can submit their work as a publication to the International Conference on Emerging Technology in Autonomous Aerial Vehicles (ETTAV) conference held in IISc Bengaluru.

• Boost your portfolio: Showcase your AI & ML skills in real-world object detection.

Queries:

For any questions, please create a discussion topic on Kaggle(Link) alternatively contact Manjunath D at manjunathd1@iisc.ac.in (with subject starting with: ETAAV-25 competition)