3-D Hybrid CNN Combined With 3-D Generative Adversarial Network for Wetland Classification With Limited Training Data
Bibliographic record
Abstract
Recently, deep learning algorithms, specifically convolutional neural networks (CNNs), have played an important role in remote sensing image classification, including wetland mapping. However, one limitation of deep CNN for classification is its requirement for a great number of training samples. This limitation is particularly enhanced when the classes of interest are spectrally similar, such as that of wetland types, and the training samples are limited. This article presents a novel approach named 3-D hybrid generative adversarial network (3-D hybrid GAN) that addresses the limited training sample issue in the classification of remote sensing imagery with a focus on complex wetland classification. We used a conditional map unit that generates synthetic training samples for only classes with a lower number of training samples to improve the per-class accuracy of wetlands. This procedure overcomes the issue of imbalanced data in conventional wetland mapping. Based on the achieved results, better classification accuracy is obtained by integrating a 3-D generative adversarial network (3-D GAN) and the CNN network of a 3-D hybrid CNN using both 3-D and 2-D convolutional filters. Experimental results on the avalon pilot site located in eastern Newfoundland, Canada, and covering five wetland types of bog, fen, marsh, swamp, and shallow water demonstrate that our model significantly outperforms other CNN models, including the HybridSN, SpectralNet, MLP-mixer, as well as a conventional algorithm of random forest for complex wetland classification by approximately 1% to 51% in terms of F-1 score.
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 | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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".