Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026.
1. Pal, Mainak; Mehrotra, Sarthak; Casari, Paolo; Chaudhuri, Subhasis; Ricci, Elisa; Banerjee, Biplab: "CLIPoint3D: Language-Grounded Few-Shot Unsupervised 3D Point Cloud Domain Adaptation. 2. Gole, Siddhant; Pal, Akash; More, Amit; Bhat, Divakar; Chaudhuri, Subhasis; Banerjee, Biplab: "Hyperbolic Prototype Learning with Uncertainty-Aware Consistency for Continual Test-Time SegmentationProceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026.
1. Kashyap, Pankhi; Singha, Mainak; Banerjee, Biplab: Bi-Modal Textual Prompt Learning for Vision-Language Models in Remote Sensing.Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2026.
1. S Gole, A Pal, A Jha, S Chaudhuri, B Banerjee. Vision-informed Semantic Text Alignment for Open-set Recognition in Remote Sensing. 2. S Choudhury, Y Salunkhe, V Rajan, S Chaudhuri, B Banerjee. X-JEPA: A Novel Joint Learning Cross-Modal Predictive Alignment Framework for Remote Sensing Image Retrieval.
GeoMeld brings together spatially aligned heterogeneous modalities—including optical imagery, SAR, elevation, land-cover products, canopy height, and structured geographic metadata—under a unified alignment protocol. A key feature of GeoMeld is its semantically grounded language supervision, generated through an agentic captioning framework that synthesizes and verifies annotations using spectral signals, terrain statistics, and geographic context.
We hope GeoMeld will serve as a useful resource for advancing multimodal foundation models in remote sensing, particularly for modality-aware representation learning, cross-modal alignment, and robust transferable geospatial representations.