PriMAT: Transfer Learning Results for Primate Tracking in the Wild

Transfer learning results for the PriMAT model, which detects and tracks nonhuman primates in the wild using bounding boxes. The dataset includes evaluation results from case studies with Assamese macaques and redfronted lemurs, and transfer tests to Barbary macaques, Guinea baboons, chimpanzees, and gorillas. The model and associated data were published by Richard Vogg under a CC-BY-4.0 license.

Use Cases

• Evaluating multi-animal tracking model performance based on results from wild primate videos.
• Assessing transfer learning efficacy for different primate species based on the described case studies.
• Training individual identification models based on the described classification branch achieving 84% accuracy for lemurs.

Strengths

• Results are derived from a model trained on only a few hundred labeled frames, as stated in the description.
• The lemur identification model component shows a reported accuracy of 84%.
• The dataset is small (5.5 KB), facilitating quick inspection.

Limitations

• The dataset is very small (5.5 KB), indicating it likely contains summary results, not raw tracking data.
• Row count and column-level documentation are unknown, limiting detailed assessment of the data structure.
• The description focuses on the model methodology; the specific contents of the XLS file are not detailed.

Provenance

Source

figshare, author Richard Vogg.

Collection Method

Likely contains evaluation metrics and results from applying the PriMAT computer vision model to video data.

Freshness

Last updated 2026-04-30 17:45:19.

Geography

Data originates from studies of primates in the wild; specific locations are not detailed.