Research map / Time series / Foundation models for time series
Satellite Image Time Series Classification With Pixel-Set Encoders and Temporal Self-Attention
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use a pixel-set encoder instead of a CNN spatial encoder (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 49.2 -> 50.9; 1.7 points | Helped |
| Add geometric parcel features (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 50.0 -> 50.9; 0.9 point | Helped |
| Use PSE-learned spectral descriptors instead of precomputed means and standard deviations (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 48.9 -> 50.9 | Helped |
| Use temporal self-attention instead of a GRU (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 47.3 -> 50.9; 3.6 points | Helped |
| Add positional encoding to the GRU input (ablation) | PSE+GRU · Agricultural parcels (AOI) | mIoU: 47.3 -> 48.7 | Helped |
| Replace temporal max-pooling with a master-query scheme (ablation) | PSE+Transformer · Agricultural parcels (AOI) | mIoU: 46.6 -> 49.5; 2.9 points | Helped |
| Use direct input embeddings as attention values rather than an intermediary value tensor (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 49.5 -> 50.9; 1.4 points | Helped |
| Use a mean query across dates rather than the last-date query (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 50.7 -> 50.9 | Helped |
| Use a mean query across dates rather than a maximum query (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 50.3 -> 50.9 | Helped |
| Reduce sampled pixels to S=16 (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 50.9 -> 50.5 | Hurt |
| Reduce sampled pixels to S=32 (ablation) | PSE+TAE · Agricultural parcels (AOI) | mIoU: 50.9 -> 50.1 | Hurt |
| Use PSE+TAE instead of CNN+GRU | Agricultural parcel classification · Agricultural parcels (AOI) | OA: 93.8 -> 94.2; 0.4 points | Helped |
Closest papers by meaning
- Efficient Representation Learning of Satellite Image Time Series and Their Fusion for Spatiotemporal Applications (AAAI 2024 · similarity 0.86)
- Exact: Exploring Space-Time Perceptive Clues for Weakly Supervised Satellite Image Time Series Semantic Segmentation (CVPR 2025 · similarity 0.86)
- Galileo: Learning Global & Local Features of Many Remote Sensing Modalities (ICML 2025 · similarity 0.84)
- TEOChat: A Large Vision-Language Assistant for Temporal Earth Observation Data (ICLR 2025 · similarity 0.82)
- TVNet: A Novel Time Series Analysis Method Based on Dynamic Convolution and 3D-Variation (ICLR 2025 · similarity 0.82)
- DiffusionSat: A Generative Foundation Model for Satellite Imagery (ICLR 2024 · similarity 0.82)
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.