Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
MorphJEPA: Morphology-Aware Latent Prediction for Hyperspectral Images
Amartya Ray, Tanmay Mandaliya, Muhammad Haris Khan, Biplab Banerjee
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add morphological open/close objectives to geometry-only supervision (ablation) | MorphJEPA · Houston 2013 (5-shot) | OA: +15.7% OA | Improved |
| add morphological objectives over geometry-only supervision (ablation) | MorphJEPA · HyRank Dioni->Loukia (5-shot) | OA: +18.3% | Improved |
| add geometric loss to morphological losses (ablation) | MorphJEPA · Houston 2013 (5-shot) | OA: 68.97 -> 73.97; std 3.72 -> 1.62 | Improved |
| add context-anchored L_reg (ablation) | MorphJEPA · Houston 2013 (5-shot) | OA: 73.97 -> 74.73 | Improved |
| add L_reg (ablation) | MorphJEPA · HyRank Dioni->Loukia (5-shot) | OA: 59.76 -> 60.37 | Improved |
| add morphological SSL pretraining to the encoder (ablation) | spatial-spectral encoder · Houston 2013 (5-shot) | OA: 31.92 -> 74.73 | Improved |
| use cross-attention fusion instead of concatenation (ablation) | MorphJEPA encoder · Houston 2013 (5-shot) | OA: 72.10 -> 74.73 | Improved |
| use cross-attention fusion instead of CBAM fusion (ablation) | MorphJEPA encoder · Houston 2013 (5-shot) | OA: 73.05 -> 74.73 | Improved |
| use MorphJEPA with 3D-CNN backbone instead of the spatial-spectral encoder (ablation) | MorphJEPA · Houston 2013 (5-shot) | OA: 64.12 vs 74.73 | Hurt |
| replace MAE with MorphJEPA | SSL baseline with same encoder · Houston 2013 (5-shot) | OA: +6.28% OA | Improved |
| replace GGCL with MorphJEPA | HSI SSL baseline · Houston 2013->2018 low-shot (5-shot) | OA: +1.20% OA | Improved |
| use LeJEPA objective instead of morphological pretext | JEPA baseline · Houston 2013 (5-shot) | OA: 24.27% OA vs 74.73 | Hurt |
Closest papers by meaning
- LESSViT: Robust Hyperspectral Representation Learning under Spectral Configuration Shift (NeurIPS 2026 · similarity 0.88)
- Unsupervised Adaptation Learning for Hyperspectral Imagery Super-Resolution (CVPR 2020 · similarity 0.87)
- Bridging Local–Global Dissonance: Learning from Compressive Measurements for Hyperspectral Reconstruction (ICML 2026 · similarity 0.87)
- EigenSR: Eigenimage-Bridged Pre-Trained RGB Learners for Single Hyperspectral Image Super-Resolution (AAAI 2025 · similarity 0.85)
- Hyperspectral Image Reconstruction via Combinatorial Embedding of Cross-Channel Spatio-Spectral Clues (AAAI 2024 · similarity 0.85)
- S2MAE: A Spatial-Spectral Pretraining Foundation Model for Spectral Remote Sensing Data (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
