Machine Learning for Remote Sensing

This workshop promotes trans-disciplinary research through diverse view-points to tackle the pressing questions of our times, such as climate change, social inequalities, biodiversity, and food security. Developing modern machine learning approaches tailored towards remote sensing data is key to investigating these problems efficiently. This third Machine Learning for Remote Sensing (ML4RS) workshop promotes this exchange by allowing researchers to present their research on environmentally and societally important applications and/or innovative methods that can have an impact in such application domains.

Topics

We solicit research papers addressing advancements in key topics in Machine Learning for Remote Sensing, such as domain adaptation, concept drift, out-of-distribution detection, evaluation using unlabeled data, model architectures for remote sensing data, semi-supervised learning, unsupervised learning, self-supervised learning, multi-fidelity data fusion, federated learning, data-centric AI, human-in-the-loop and active learning, machine learning for time series, methods for learning from limited labeled data (e.g., few-shot learning, meta-learning), new benchmark datasets involving remote sensing data, geographic equity, and fairness.

We welcome applications related to sustainable development, societal needs, planetary exploration, and more, including but not limited to agriculture and food security, forestry, biodiversity and species distribution modeling, natural hazards and disasters, and other societal and environmental questions.

Papers can present methodological innovations designed towards a particular problem or can also apply existing methods when they highlight new perspectives or limitations on existing methods or the broader environmental and societal applicability through a particular dataset or impact area.

Call for Papers

Important Dates

Bonus Track

We invite content for a “bonus track” of our workshop to better broadcast projects, progress, and announcements across the diverse orbits of work on machine learning for remote sensing. In particular we welcome the submission of one-slide presentations and live demos.

The deadline for the bonus track is April 23, 2025 (Anywhere on Earth).

Accepted slides will be auto-played during the poster sessions and breaks and linked from the workshop site. Demos will be presented in-person during the event subject to venue space and equipment.

Travel Support

Thanks to our sponsors, we have limited funding available to support the travel of students to attend the workshop held at ICLR 2024. Awards are based on merit with additional consideration based on need and travel distance. Priority will be given to those whose papers are accepted for presentation at the workshop.

The deadline for submitting this application was March 28th, 11:59pm Anywhere on Earth.

Paper Format

This year, we have two tracks:

  1. a workshop paper track (4-pages) and
  2. a tiny paper track (2-pages)

Please indicate the track in the submission form.

Workshop Paper Track

We invite short papers describing new and ongoing/in progress research of 4 pages. To prepare your submission, please use the LaTeX style files for the ML4RS workshop ICLR 2025 that provides further detail on the paper structure. Paper reviews will be double blind. When submitting your manuscript, make sure you do not include any personally-identifying information such as author names or GitHub links which would de-anonymize the submission.

Page limits do not include references, which are unlimited. The workshop papers will be non-archival and dual submission is allowed where permitted by third parties. After the workshop selected papers will be invited to submit to a special stream in the IEEE Geoscience and Remote Sensing Letters (please note that this will incur a - fast - additional review phase and additional publication fees).

Machine Learning for Remote Sensing is non-archival and thus dual submission is allowed where permitted by third parties.

Tiny Paper Track

Following the ICLR 2023 and 2024 initiatives, we are delighted to implement Tiny Papers in the workshop as part of the DEI initiative at ICLR 2025!

The objectives of the “Tiny Papers” track at ICLR workshops are threefold:

Despite the wide range of topics that could benefit from the Tiny Papers format, we focus the type of submissions to be the following:

The guidelines and restrictions for this Tiny Papers track are:

Write a paper with the same structure as seen in full papers, but with at most 2 pages of main text. References, Appendices do not count towards the page limit, but we do not recommend dissecting a regular length paper into a tiny paper by moving the majority of the content into appendices. It is recommended to have 1-2 figures, 1 table, and 4-5 sections. But authors are free to use their own structure.

Paper Submission

Please submit your paper before the deadline (see important dates!) via CMT.

You can still edit your submission until the deadline. You can also choose to submit a supplementary pdf alongside the 4-page paper. However, the reviewers are not required to look at the supplementary files.

🗓 Schedule

Event times shown in the schedule are local times in Singapore

Time Topic
9:00–9:10 Opening Remarks
9:10–9:40 Keynote 1
Stefano Ermon: Topic TBA
9:40–10:00 Invited Talk 1
Hao Li : Topic TBA
10:00–10:30 Coffee Break
10:30–10:50 Invited Talk 2
Soo Chin Liew : Topic TBA
10:50–11:20 Oral Talks 1
- A Recipe for Improving Remote Sensing VLM Zero Shot Generalization
- Using Multiple Input Modalities Can Improve Data-efficiency for ML with Satellite Imagery
11:20–12:00 Poster Session #1
- Self-supervised Object Detection in Synthetic Aperture Radar Images
- Zero-Shot Tree Detection and Segmentation from Aerial Forest Imagery
- A Recipe for Improving Remote Sensing VLM Zero Shot Generalization
- Leveraging Satellite Imagery for Childhood Poverty Estimation
- Interactive Few-shot Online Adaptation for User-Guided Segmentation in Aerial Images
- PyViT-FUSE: A Foundation Model for Multi-Sensor Earth Observation Data
- Monitoring Illicit Rare Earth Mining in Myanmar via Self-Supervised Learning
- DR-SCAN: An Interpretable Dual-Branch Residual Spatial and Channel Attention Network for Remote Sensing and Geoscience Image Super-Resolution
- DEAL-YOLO: Drone-based Efficient Animal Localization using YOLO
- Using Multiple Input Modalities Can Improve Data-Efficiency for ML with Satellite Imagery
- Robustness to Geographic Distribution Shift using Location Encoders
- An Analysis of Multimodal Large Language Models for Object Localization in Earth Observation Imagery
- LIGHTHOUSE: Fast and Precise Distance to Shoreline Calculations from Anywhere on Earth
12:00–14:00 Lunch Break
14:00–14:30 Keynote 2
Xiaoxiang Zhu : Topic TBA
14:30–15:00 Oral Talks 2
- Balancing Quantity and Representativeness in Constrained Geospatial Dataset Design
- Do Satellite Tasks Need Special Pretraining?
15:00–15:30 Coffee Break
15:30–15:50 Invited Talk 3
Konstantin Klemmer: Topic TBA
15:50–16:20 Oral Talks 3
- Tackling Few-Shot Segmentation in Remote Sensing via Inpainting Diffusion Model
- OSDMamba: Enhancing Oil Spill Detection from Synthetic Aperture Radar Images Using Selective State Space Model
16:20–17:00 Poster Session 2
- Onboard Terrain Classification via Stacked Intelligent Metasurface-Diffractive Deep Neural Networks from SAR Level-0 Raw Data
- Tackling Few-Shot Segmentation in Remote Sensing via Inpainting Diffusion Model
- Metadata, Wavelet, and Time Aware Diffusion Models for Satellite Image Super Resolution
- Assessing SAM for Tree Crown Instance Segmentation from Drone Imagery
- LandsatQuake: A Large-Scale Dataset for Practical Landslide Detection
- OSDMamba: Enhancing Oil Spill Detection from Synthetic Aperture Radar Images Using Selective State Space Model
- Efficient Land-Cover Image Classification via Mixed Bit-Precision Quantization
- Predicting Out-of-Domain Performance under Geographic Distribution Shifts
- Balancing Quantity and Representativeness in Constrained Geospatial Dataset Design
- SAMSelect: A Spectral Index Search for Marine Debris Visualization Using Segment Anything
- Self-Supervised Representation Learning on Remote Sensing Pixel Time Series with Patch-Based Masking
- Large Language Models for Captioning and Retrieving Remote Sensing Images
- Do Satellite Tasks Need Special Pretraining?
17:00–17:20 Best Paper Awards, Closing Remarks

Keynote Speakers

Stefano Ermon Xiaoxiang Zhu

Invited Speakers

Hao Li Soo Chin Liew Konstantin Klemmer

Organizers

Program Committee

We thank the program committee for being available to review the submitted papers!

Sponsorship

We are looking for sponsors to support travel grants, best paper awards, and other workshop activities. If you or your organization is interested in becoming a workshop sponsor, please contact the organizing team using the email below.

Sponsors

We are looking for more sponsors. Please contact us per email if you would like to support this workshop.

Gold Sponsors

Hyperplan

Silver Sponsors

SIG GIS Allen AI

Contact

For questions or information about the Machine Learning for Remote Sensing workshop at ICLR 2025 please contact ml4rs_iclr25@googlegroups.com.