Machine vision automated species identification scaled towards production levels
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Abstract Computer‐automated identification of insect species has long been sought to support activities such as environmental monitoring, forensics, pest diagnostics, border security and vector epidemiology, to name just a few. In order to succeed, an automated identification programme capable of addressing the needs of the end user should be able to classify hundreds of taxa, if not thousands, and is expected to distinguish closely related and hence morphologically similar species. However, it remains unknown how automated identification methods might handle an increase in data quantity, be it in reference imagery or taxonomic diversity. We sought to test the scalability of an automated identification method in terms of the number of reference specimens used to train the classifier and the number of taxa into which the classifier should assign unknown specimens. Is there an optimal number of reference images, where the cost of acquiring more images becomes greater than the marginal increase in identification success? Does increasing taxonomic diversity affect identification success, whether negatively or positively? In order to test the scalability of the automated insect identification enterprise, we used a sparse processing technique and support vector machine to test the largest dataset to date: 72 species of fruit flies ( D iptera: T ephritidae) and 76 species of mosquitoes ( D iptera: C ulicidae). We found that: (i) machine vision methods are capable of correctly classifying large numbers of closely related species; (ii) when the misclassification of a specimen occurs at the species level, it is often classified in the correct genus; (iii) classification success increases asymptotically as new training images are added to the dataset; (iv) broad taxon sampling outside a focal group can increase classification success within it.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 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 it