Leveraging synthetic data produced from museum specimens to train adaptable species classification models
Bibliographic record
Abstract
Computer vision has increasingly shown potential to improve data processing efficiency in ecological research. However, training computer vision models requires large amounts of high-quality, annotated training data. This poses a significant challenge for researchers looking to create bespoke computer vision models, as substantial human resources and biological replicates are often needed to adequately train these models. Synthetic images have been proposed as a potential solution for generating large training datasets, but models trained with synthetic images often have poor generalization to real photographs. Here we present a modular pipeline for training generalizable classification models using synthetic images. Our pipeline includes 3D asset creation with the use of 3D scanners, synthetic image generation with open-source computer graphic software, and domain adaptive classification model training. We demonstrate our pipeline by applying it to skulls of 16 mammal species in the order Carnivora. We explore several domain adaptation techniques, including maximum mean discrepancy (MMD) loss, fine-tuning, and data supplementation. Using our pipeline, we were able to improve classification accuracy on real photographs from 55.4% to a maximum of 95.1%. We also conducted qualitative analysis with t-distributed stochastic neighbor embedding (t-SNE) and gradient-weighted class activation mapping (Grad-CAM) to compare different domain adaptation techniques. Our results demonstrate the feasibility of using synthetic images for ecological computer vision and highlight the potential of museum specimens and 3D assets for scalable, generalizable model training.
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 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| 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".