.png)
Computer vision models
WildObs develops computer vision models tailored to Australian species and environments to support accurate, scalable annotation of wildlife camera-trap images. These models are designed for research and monitoring workflows, combining automated detection with human validation.
A multi-scale modelling approach
WildObs uses a multi-scale approach, developing national models for broad applicability alongside region-specific models tuned for local accuracy.
National wildlife computer vision model
The WildObs national computer vision model is being developed to support species detection and classification across Australia. It aims to provide broad species coverage and consistent baseline predictions across regions, projects, and time, supporting large-scale and cross-regional analysis.
The model is under active development and will be released progressively as new tagged image data becomes available.
Region-specific wildlife computer vision models
WildObs provides region-specific computer vision models optimised for particular environments and species assemblages. Trained on regionally relevant data, these models focus on defined sets of species labels and survey conditions to improve accuracy for local research and monitoring programs.
Regional models complement the national model by delivering higher performance in specific ecological contexts.
Models ready to use
Explore region-specific computer vision models currently available through WildObs, with additional models in development to expand coverage across Australia.
WildObs QLD Wet Tropics model
WildObs QLD Wet Tropics model is a region-specific computer vision model developed for wildlife camera-trap imagery from the Australian Wet Tropics. Trained on a curated subset of tagged images from the WildObs repository and fine-tuned using SpeciesNet, the model supports accurate species classification for locally common fauna and demonstrates high performance, with evaluation results frequently exceeding 95% F1-score.
• Region : Australian Wet Tropics
• Source: WildObs Tagged Image Repository
• Base model: SpeciesNet
• Underlying dataset: 454 camera deployments, ~2.18 million images, 121 recorded species
• Training data: ~30,000 curated and labelled images
• Alectura lathami or Australian brush-turkey
• Bos taurus or domestic cattle
• Canis familiaris or domestic dog
• Casuarius casuarius or southern cassowary
• Felis catus or domestic cat
• Heteromyias cinereifrons or grey-headed robin
• Homo sapiens or human
• Hypsiprymnodon moschatus or musky rat kangaroo
• Megapodius reinwardt or orange-footed scrubfowl
• Orthonyx spaldingii or northern chowchilla
• Perameles nasuta or long-nosed bandicoot
• Sus scrofa or wild boar
• Thylogale stigmatica or red-legged pademelon
• Uromys caudimaculatus or giant uromys
• Wallabia bicolor or swamp wallaby
Coming soon
WildObs is actively developing additional region-specific computer vision models to expand coverage across Australia’s diverse ecosystems. These models are being trained and evaluated using curated tagged image data and will be released progressively.
WildObs QLD K’gari model
A region-specific computer vision model optimised for wildlife camera-trap imagery from K’gari (Fraser Island). The model is designed to support species commonly observed in coastal and island ecosystems and will complement existing regional and national models.
WildObs Australian National model
A national computer vision model under development for wildlife camera-trap imagery across Australia. This model is designed to recognise a broad range of species and environments, supporting consistent annotation and scalable research and monitoring nationwide.
Working together at national scale
WildObs partners with leading Australian research and conservation organisations to support large-scale monitoring, data sharing, and ecological research across Australia.
WildObs is a co-investment partnership with the Australian Research Data Commons (ARDC) through the Planet Research Data Commons (DOI: 10.3565/bvg2-b035). The ARDC is enabled by the Australian Government’s National Collaborative Research Infrastructure Strategy (NCRIS).
_edited.png){kind=logo}
