JEPA · 2026-11-05
ChangeJEPA: Predict Before You Compare for Remote Sensing Change Detection
Buddhi Wijenayake, Dhananjaya Jayasundara, Vishal M. Patel
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
ChangeJEPA reformulates remote sensing change detection as latent post-state prediction. Given a pre-change image and partial post-change evidence, it predicts a full post-change latent representation under two complementary conditions (prior vs updated). Change is inferred from disagreement among pre representation, observed post target, and predicted latent states, aiming to suppress nuisance variation while keeping semantic consistency.
Why it matters
The authors argue that Siamese feature comparison is ambiguous because encoder discrepancies can reflect both true change and nuisance factors (illumination, seasonal/radiometric/geometric effects, registration errors). By aligning prior and updated predictions to a shared post-state latent target and using their disagreement as evidence, ChangeJEPA provides a semantically aligned space for change segmentation.
Method
- Bitemporal JEPA: predict the complete post-change latent representation from the full pre-change image plus partial post-change context, using two branches (prior with 0% evidence; updated with p% visible evidence).
- Joint-embedding predictive loss aligns both hypotheses with the same complete post-change target latent representation, encouraging semantic consistency and suppressing transient appearance variation.
- Change evidence is computed by multiple pairwise latent comparisons (between pre, observed post, prior prediction, and updated prediction), aggregated to produce a dense change map via a decoder and morphological refinement.
Limitation
The controlled-nuisance experiment on LEVIR-CD+ alters pre-event appearance (without changing I2 or reference masks), and the authors report that accuracy deteriorates as degradation severity increases, showing sensitivity to substantial degradation of the pre-event observation.
Abstract (from arXiv)
Remote sensing change detection is typically formulated as Siamese feature comparison, where two images are encoded independently and a decoder determines which discrepancies correspond to meaningful change. This is challenging because the encoders preserve both true scene changes and nuisance variations such as illumination, seasonal, radiometric, and geometric differences. We introduce ChangeJEPA, a bitemporal joint-embedding predictive architecture that reformulates change detection as latent post-state prediction. Given a pre-change image and partial post-change evidence, ChangeJEPA predicts the full post-change representation under two complementary conditions: a prior hypothesis derived from the pre-change scene and an updated hypothesis conditioned on observed post-change context. A joint-embedding predictive loss aligns both hypotheses with the target latent representation, encouraging semantic consistency while suppressing transient appearance variations. Change is then inferred from the disagreement among the pre-change representation, the observed post-change target, and the predicted latent states, capturing complementary evidence of change, prior surprise, contextual update, and prediction reliability. Across SYSU-CD, LEVIR-CD+, and CDD, ChangeJEPA consistently outperforms its decoder matched Siamese baseline by up to 1.51 IoU points and achieves the best IoU and F1-score on all three benchmarks, demonstrating the advantage of latent post-state prediction over direct feature comparison. Page: https://buddhi19.github.io/ChangeJEPA
Related papers
- T-JEPA: A Temporal Joint-Embedding Predictive Architecture for Learning Better Remote Sensing Representations
- RS-JEPA: Remote Sensing Joint-Embedding Predictive Architecture with Scan-Path Views for Cross-Sensor Image Retrieval
- CAPE-JEPA - Cross-modal Adversarial Probabilistic Embedding JEPA
- Latent-Foresight: End-to-End Learning Predictable Representations for Latent World Models
- S2S-JEPA: Predicting the Predictable at Subseasonal-to-Seasonal Timescales
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-11-05 · Indexed by Pub-AI: 2026-11-05 UTC
ChangeJEPA: Predict Before You Compare for Remote Sensing Change Detection
First announced by arXiv: 2026-11-05 · Original record ↗
- arxiv
- 2611.05570
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-11-05 · First detected: 2026-11-05 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.05570v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
Remote sensing change detection is typically formulated as Siamese feature comparison, where two images are encoded independently and a decoder determines which discrepancies correspond to meaningful change. This is challenging because the encoders preserve both true scene changes and nuisance variations such as illumination, seasonal, radiometric, and geometric differences. We introduce ChangeJEPA, a bitemporal joint-embedding predictive architecture that reformulates change detection as latent post-state prediction. Given a pre-change image and partial post-change evidence, ChangeJEPA predicts the full post-change representation under two complementary conditions: a prior hypothesis derived from the pre-change scene and an updated hypothesis conditioned on observed post-change context. A joint-embedding predictive loss aligns both hypotheses with the target latent representation, encouraging semantic consistency while suppressing transient appearance variations. Change is then inferred from the disagreement among the pre-change representation, the observed post-change target, and the predicted latent states, capturing complementary evidence of change, prior surprise, contextual update, and prediction reliability. Across SYSU-CD, LEVIR-CD+, and CDD, ChangeJEPA consistently outperforms its decoder matched Siamese baseline by up to 1.51 IoU points and achieves the best IoU and F1-score on all three benchmarks, demonstrating the advantage of latent post-state prediction over direct feature comparison. Page: https://buddhi19.github.io/ChangeJEPA
