Dual Image-Based Deep Learning System for Predicting Soil Texture and Soil Organic Matter Content
Bibliographic record
Abstract
Soil texture, a key factor in soil health and agricultural productivity, has traditionally been measured using time-consuming laboratory methods. This study takes a novel approach by exploring the prediction of soil texture and soil organic matter (SOM) using an innovative image-based data fusion technique. This method, which combines images captured at two different spatial scales: 8.85 um per pixel and 0.70 um per pixel, was implemented in a low-cost prototype that incorporates a camera and a microscope, both controlled through a Raspberry Pi 4 board. The detailed images of soil samples captured by this system were used as inputs for models derived from the ConvNeXt architecture, a robust convolutional neural network (CNN) for image processing.The Insitut de Recherche et de Development en Agroenvironnement (IRDA), a respected institution in the field, carefully selected these samples to represent Quebec’s agricultural soils, encompassing a wide range of physical properties. 793 samples from 272 sites and 68 soil series were included, with texture and SOM measurements performed using the hydrometer method and loss on ignition, respectively. A cross-validation technique was employed to ensure robust model evaluation, approximating the expected performance and stability of the models. The results of this study confirm the robustness of the model, with the combination of camera and microscope images significantly enhancing prediction performance, particularly for sand and silt fractions. The best-performing model achieved RMSE values of 8.36% for sand, 5.77% for silt, 6.04% for clay, and 1.53% for SOM. Furthermore, the study suggests that different spatial scales and image resolutions are critical factors influencing model performance, thereby, opening new avenues for future research. This study validates the potential of an image-based approach as a soil texture and SOM measurement method. It highlights the benefits that data fusion of information at different spatial scales can have in enhancing model performance and robustness
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 | 0.001 | 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".