Deep Learning and Synthetic Imagery for Migratory Bird Species Identification Using Drones
Bibliographic record
Abstract
Recent technology advancements have caused a significant increase in popularity and availability of uninhabited aerial systems (UAS).Commercially available platforms are often equipped with powerful sensors, and have become a tool for researchers in many fields.In ornithology, using drones for surveying and population monitoring of different bird species is a relatively lower cost and higher time efficiency approach, when compared to ground based alternatives.In this thesis work, the use of UAS in the study of migratory shorebird species in Canada is explored with the development of computer vision applications.More specifically, a deep learning classification model is trained to identify the presence of birds of a given species in an image.The species selected for the first iteration of the application is the Canada geese, which is not one of the migratory shorebird species of interest, but is used as a proxy due to its availability and similar melanin colours.Images were collected from UAS in several field trips for the development of the vision models, and realistic physical models of the species of interest were used, considering the challenges of collecting data from live animals at that point in the project.Due to the specific nature of the problem being solved, the constructed datasets available for training the models did not have the volume that deep learning models typically require.To address this data scarcity issue, the datasets used were augmented with synthetic data generated with a Blender software add-on, with realistic models of the birds.For evaluation of the quality of the artificially generated images, a novel measure is developed, which is based on the reconstruction error of a convolutional autoencoder and a measure of image complexity.The synthetic image quality measure, named reconstruction error based similarity (RES), showed better results in controlled environments when compared to a popular alternative iii List of TablesHyperparameter search candidate values for the convolutional autoencoder.Random combinations of values are sampled during Bayesian search optimization. . . . .
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) | 0.000 | 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 | 0.001 |
| 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".