Call for papers

Call for papers 

Submission: OpenReview platform https://openreview.net/group?id=thecvf.com/WACV/2027/Workshop/CV4EO

Author guidelines (follow WACV main conference): https://wacv.thecvf.com/Conferences/2027/AuthorReviewerGuides

• Author kit (.zip)

We invite two types of submissions:

• Full papers (up to 8 pages, including figures and tables) presenting mature, original research on computer vision for earth observation methods and applications. Submissions should follow WACV’s formatting and double-blind review guidelines, with accepted papers to appear in the WACV 2027 Workshop Proceedings.
  
• Short papers (up to 4 pages, excluding References/Appendix) including position papers, abstracts on work-in-progress, industrial applications demos, or descriptions of workflows and best practices for CV4EO. Accepted short papers will be part of the lightning talks and poster sessions, but will not be in the Proceedings.
• Workshop takeaways and accepted short papers will be invited for compilation into an extended journal article submission to IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing (JSTARS).

Topics include, but are not limited to:

• Foundation models for Earth observation applications, including architectures, self-supervised learning strategies, and approaches integrating spatiotemporal characteristics or geospatial metadata.
• Earth embeddings and world models: reusable embedding fields as analysis-ready products, retrieval and similarity search at scale, and generative or predictive models of Earth surface dynamics.
• Data fusion and multimodal learning: harmonization, tokenization, and integration of diverse modalities (e.g., optical multispectral, SAR, LiDAR) and diverse data types (e.g., raster, vector, text).
• Agentic AI in remote sensing. Integration of LLMs with computer vision techniques for reasoning and querying of Earth observation applications.
• 3D and geometric vision for EO: multi-view stereo, height/DSM estimation, and cross-view matching,  neural rendering
• Practical computer vision and machine learning for low resource EO applications. Energy-efficient architectures, compression, parameter-efficient adaptation, on-board/edge deployment.
• Benchmarking and evaluation protocols for CV4EO models: uncertainty quantification, trustworthiness, generalization, explainability, and application-oriented workflows.
• Ethical considerations (bias, privacy, fairness, accountability, reproducibility) and open challenges, case studies, and end-user application perspectives.
• Visual query answering, object detection, semantic/instance segmentation, classification techniques for applications in humanitarian assistance, precision agriculture, sustainable development goals, environment/wildlife monitoring etc.
• Emerging advances: quantum machine learning for Earth observation, diffusion models for remote sensing data synthesis etc

CV4EO scope:

Some defining features of CV4EO as compared to other CV domains are:

1. Planetary/Environmental Purpose (Mandatory): The core goal must be the characterization, monitoring, or modeling of Earth’s (or other planetary bodies) physical, chemical, or biological states.
2. Geospatial & Spatial Awareness: The data or methodology considers geographic coordinates, structural scales, or localized planetary context.
3. Multi-Modal Sensing: The integration of data beyond RGB, such as LiDAR, SAR, hyperspectral imaging, or time-series sensor fusion.

For example, a work discussing environmental mapping using multimodal data would be within scope, even if it doesn’t include overhead imagery. In contrast, a work using RGB-only, ground-level imagery without any geospatial awareness/scaling, would likely be out of scope. A work targeting self-driving cars with multimodal data would also likely be out of scope (see item 1).

About archival (full) papers:

• WACV Workshop proceedings will appear in IEEE Xplore but will be indexed separately from the main conference
• Archival-paper authors must attend and present in person under WACV 2027 policy.
• All full papers will undergo peer-review by minimum 3 reviewers. : Given the growth in the number of paper submissions, we expect all authors to be willing to serve as reviewers if asked to do so.
• Authors will be required to have an up-to-date OpenReview profile

Dates: