Installation¶

Prerequisites¶

Python 3.8+ (3.10+ recommended)
Optional: NVIDIA GPU with CUDA 12.1 for 10–50x speedup

pip¶

pip install PytorchWildlife

Conda¶

conda create -n pytorch-wildlife python=3.10 -y
conda activate pytorch-wildlife
pip install PytorchWildlife

Windows users: Use the Anaconda Prompt if using Anaconda, otherwise use PowerShell.

GPU Setup¶

Check if CUDA is available¶

import torch
print(torch.cuda.is_available())

Install GPU-enabled PyTorch (CUDA 12.1)¶

If torch.cuda.is_available() returns False on a CUDA machine:

pip uninstall torch torchvision torchaudio
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
pip install PytorchWildlife

Ubuntu — OpenCV dependency¶

sudo apt-get update
sudo apt-get install -y python3-opencv

macOS — ffmpeg for video decoding¶

brew install ffmpeg

Windows¶

See the Windows installation guide for a step-by-step walkthrough.

Verify Installation¶

from PytorchWildlife.models import detection as pw_detection

model = pw_detection.MegaDetectorV6()
print("PyTorch-Wildlife loaded successfully.")

Model weights download automatically on first use.

Try Without Installing¶

Hugging Face demo — upload images in your browser
Google Colab notebook — free cloud GPU

Docker¶

docker pull andreshdz/pytorchwildlife:1.0.2.3
docker run -p 80:80 andreshdz/pytorchwildlife:1.0.2.3 python demo/gradio_demo.py

Jupyter Notebooks¶

To use the demo notebooks with Jupyter:

conda install ipykernel
python -m ipykernel install --user --name pytorch-wildlife --display-name "Python (PytorchWildlife)"

Then select the Python (PytorchWildlife) kernel when running notebooks.