Skip to content

Model Zoo

The Polli Linnaeus Model Zoo provides access to pre-trained models for taxonomic classification. These models can be used for direct inference or as a basis for further research and fine-tuning. Our initial set of models focuses on North American taxa, built using the mFormerV1_sm architecture.

All models listed below are scheduled for release on Hugging Face Hub (under polli-caleb/linnaeus) in early July. Links will be updated upon release.

Polli Linnaeus - North American Aves (Birds)

  • Description: A model for classifying bird species found in North America.
  • Architecture: mFormerV1_sm
  • Hugging Face Link: https://huggingface.co/polli-caleb/linnaeus-aves-mformerV1_sm-v1 (Link will be active upon release in early July)
  • Taxonomic Scope: Birds (Aves) of North America.
  • Key Metrics: (To be determined upon final evaluation)
  • Expected Input:
    • Image: 224x224 pixels, 3 channels (RGB)
    • Metadata: None expected by default.
  • Inference Guide: [Link to ../inference/running_inference_with_pretrained_models.md]

Polli Linnaeus - North American Amphibia

  • Description: A model for classifying amphibian species (frogs, salamanders, etc.) found in North America.
  • Architecture: mFormerV1_sm
  • Hugging Face Link: https://huggingface.co/polli-caleb/linnaeus-amphibia-mformerV1_sm-v1 (Link will be active upon release in early July)
  • Taxonomic Scope: Amphibians (Amphibia) of North America.
  • Key Metrics: (To be determined upon final evaluation)
  • Expected Input:
    • Image: 224x224 pixels, 3 channels (RGB)
    • Metadata: None expected by default.
  • Inference Guide: [Link to ../inference/running_inference_with_pretrained_models.md]

Polli Linnaeus - North American Reptilia

  • Description: A model for classifying reptile species (snakes, lizards, turtles, etc.) found in North America.
  • Architecture: mFormerV1_sm
  • Hugging Face Link: https://huggingface.co/polli-caleb/linnaeus-reptilia-mformerV1_sm-v1 (Link will be active upon release in early July)
  • Taxonomic Scope: Reptiles (Reptilia) of North America.
  • Key Metrics: (To be determined upon final evaluation)
  • Expected Input:
    • Image: 224x224 pixels, 3 channels (RGB)
    • Metadata: None expected by default.
  • Inference Guide: [Link to ../inference/running_inference_with_pretrained_models.md]

Polli Linnaeus - North American Primary Terrestrial Arthropoda

  • Description: A model for classifying common terrestrial arthropods (insects, spiders, etc.) found in North America.
  • Architecture: mFormerV1_sm
  • Hugging Face Link: https://huggingface.co/polli-caleb/linnaeus-arthropoda-mformerV1_sm-v1 (Link will be active upon release in early July)
  • Taxonomic Scope: Primary Terrestrial Arthropods (e.g., Insecta, Arachnida) of North America.
  • Key Metrics: (To be determined upon final evaluation)
  • Expected Input:
    • Image: 224x224 pixels, 3 channels (RGB)
    • Metadata: None expected by default.
  • Inference Guide: [Link to ../inference/running_inference_with_pretrained_models.md]

Polli Linnaeus - North American Angiospermae

  • Description: A model for classifying flowering plant species found in North America.
  • Architecture: mFormerV1_sm
  • Hugging Face Link: https://huggingface.co/polli-caleb/linnaeus-angiospermae-mformerV1_sm-v1 (Link will be active upon release in early July)
  • Taxonomic Scope: Flowering Plants (Angiospermae) of North America.
  • Key Metrics: (To be determined upon final evaluation)
  • Expected Input:
    • Image: 224x224 pixels, 3 channels (RGB)
    • Metadata: None expected by default.
  • Inference Guide: [Link to ../inference/running_inference_with_pretrained_models.md]

Polli Linnaeus - North American Mammalia

  • Description: A model for classifying mammal species found in North America.
  • Architecture: mFormerV1_sm
  • Hugging Face Link: https://huggingface.co/polli-caleb/linnaeus-mammalia-mformerV1_sm-v1 (Link will be active upon release in early July)
  • Taxonomic Scope: Mammals (Mammalia) of North America.
  • Key Metrics: (To be determined upon final evaluation)
  • Expected Input:
    • Image: 224x224 pixels, 3 channels (RGB)
    • Metadata: None expected by default.
  • Inference Guide: [Link to ../inference/running_inference_with_pretrained_models.md]

Additional models covering more taxa and geographic regions are planned for future releases.