Full-length papers
Accepted papers that may opt in to archival proceedings; presented as spotlight talks and posters.
Computer Vision/CORE Unranked
TerraBytes II 2026
ECCV 2026 Workshop TerraBytes II
TerraBytes II is a workshop at ECCV 2026 focused on global datasets and models for Earth observation, addressing challenges in data curation, archiving, and representation learning for satellite and remote sensing data. It brings together researchers to present accepted papers on benchmarks, generative models, image registration, and downstream applications, following a successful first edition at ICML 2025.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Short papers
Non-archival papers presented as posters.
Research areas in scope
01
Datasets & Benchmarks
SwissCrop25: A National Multi-Year Benchmark for Operational Crop MappingGEOID-Flood: A Large-Scale Multi-Modal Benchmark Dataset for Flood SegmentationDisasterInsight: A Building-Centric Benchmark for Evaluating Vision–Language Models in Disaster ResponsePhindMe: A Difficulty-Stratified PhiSat-2 Benchmark for Resource-Constrained Satellite Image GeoreferencingSen2Air and SentX-Air: Datasets and Benchmarks for Air Pollutant EstimationOceanDepths: A Global Dataset of Paired Subsurface and Surface Ocean ObservationsEmbeddings based Anomaly Detection for Cleaning Global Crop Type Reference DatasetsMELBA: Multi-temporal benchmark for land-cover, building density, and gold-panning tasks
02
Generative and Self-Supervised Learning
VLM2GeoVec: Toward Universal Multimodal Embeddings for Remote SensingTerraDiT: Point-Conditioned Diffusion Transformer for Satellite Image SynthesisDual-Form Foundation Model for Online Earth MonitoringHelios: An efficient family of self supervised remote sensing modelsFar from the Crowd: Scalable Self-Supervised Learning via Geographic Isolation
03
Image Registration & Restoration
Spatial-Frequency Gated Swin Transformer for Cross-Sensor Remote Sensing Super-ResolutionRobust Satellite RPC Refinement via Bundle Adjustment with Season-Invariant CorrespondencesRAPU: Real ALS Point Upsampling Dataset and Geometry-Appearance Conditioned Diffusion Model for Airborne LiDAR Scene UpsamplingALIGNet: Zero-Shot Multimodal Remote Sensing Image Registration with Sensor-Like AugmentationSeasonStereo: Robust Dense Stereo Matching for Multi-Date Satellite Imagery via Generative AIAir Quality Downscaling with Station-Guided Pseudo-Supervision
04
Classification, Segmentation and Retrieval
Locally Consistent Transductive Information Maximization for Few-Shot Remote Sensing Scene ClassificationFew-Shot Open-Vocabulary Remote Sensing Segmentation via Textual InversionRevisiting Semantic Segmentation in Earth Observation: Pixel-Level Balanced Contrastive Learning with Graph AttentionDistilled Roads: Generalisable Road Network Extraction Across Sensors, Resolutions, and RegionsSkySpector: Query-Centric Multi-Positive Retrieval for Earth Observation
Policies worth checking twice
- Submissions are closed.
- Accepted full-length papers may opt in to archival proceedings; short papers are non-archival.
- Travel grant applications are closed and prioritized for early career scientists who are authors of accepted papers and would otherwise be unable to attend.
- Author registration closed on Monday 10 August 2026.
- All accepted papers are presented in poster sessions, with selected papers given spotlight talks.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026