SpectroChangeNet Model for Change Detection in Synthetic Aperture Radar
Bibliographic record
Abstract
Change detection using Synthetic Aperture Radar (SAR) has gained increased attention from the research community.The majority of existing change detection models have focused primarily on detection performance, with limited attention given to computational time and memory usage.To address this concern and achieve better change detection performance, a new deep learning model named SpectroChangeNet is introduced in this paper.The SpectroChangeNet model utilizes feature vectors from both the frequency and spatial domains to improve change detection results.Specifically, in the frequency domain, discrete coefficients are extracted from the SAR images using the Discrete Cosine Transform (DCT).In the spatial domain, the Multi-Region Convolution (MRC) module extracts deep feature vectors from the input images.The combined discrete coefficients and deep feature vectors are then used to classify uncertain pixels as changed or unchanged using the softmax function.In addition, a hybrid loss function is integrated with the MRC module to reduce the model's computational time and memory usage.The focal loss efficiently down-weights well-classified samples to concentrate more on hard or mis-classified uncertain samples.Further, the Mean Absolute Error (MAE) ensures that the gradient updates remain balanced and smooth.This model achieved 86.98%, 96.20%, and 94.28% of Kappa Coefficient (KC), and 96.36%, 98.92%, and 98.52% of Percentage of Correct Classification (PCC) on the Yellow River, Sulzberger, and Ottawa datasets.The achieved change detection results are superior compared to the existing models.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".