Deep learning for pockmark detection: Implications for quantitative seafloor characterization
Bibliographic record
Abstract
Occurring globally, pockmarks are seafloor depressions associated with seabed fluid escape. Pockmark ubiquity and morphologic heterogeneity result in an irregular seafloor that can be difficult to quantitatively describe. To address this challenge, we test the hypothesis that deep-learning based object detection and segmentation can be used to develop data-driven models for pockmark identification and characterization. This study describes the development, testing, and deployment of eight separate deep learning-based pockmark detection models using publicly available, gridded bathymetric data from the Belfast Bay, Maine, USA, Blue Hill Bay, Maine, USA, and Passamaquoddy Bay, New Brunswick, Canada estuarine pockmark fields. The models tested include three types of convolutional neural network architectures, as well as a generative adversarial network. We find that the data-driven models consistently resolve pockmarks from the background seafloor, allowing for quick and accurate delineation of pockmarks in a variety of seabed habitats. With these delineations we examine and compare the morphology of the muddy estuarine pockmark fields. We then compare these morphometric results to pockmark fields in two distinct settings, the sandy German Bight and the Aquitaine continental slope. We find that in all the pockmark fields a power law relationship, generally, exists between pockmark area and pockmark depth, though this relationship deteriorates with the smallest pockmarks, suggesting that there may be a minimum size needed for geomorphic stability. These results show that the training data and trained models developed here can be applied for quick detection and characterization of pockmarks where other high-resolution bathymetry is available, demonstrating the value of data-driven detection models for characterizing morphologically complex seafloors. Last, the morphologic characteristics of pockmarks identified in this study will aid future studies in relating pockmark size to environmental characteristics like seabed sediment texture and regional gradient.
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 | 0.003 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| 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".