See the PyPI project for details.
More training data means better the model accuracy (in theory). However, my ability to curate training data is limited. If you'd like to contribute, please upload your curated files into the appropriate subdirectories.
Not happy with my models? Think you can do better? Great! You can train your own models with training/train.py:
Usage: train.py [OPTIONS]
Options:
--focused-training-data-directory TEXT
The directory containing training images for
the focused model, organized into 'Yes' and
'No' subdirectories. [required]
--models-directory TEXT The directory in which to store the models.
[required]
--portrait-training-data-directory TEXT
The directory containing training images for
the portrait model, organized into 'Yes' and
'No' subdirectories. [required]
--training-time-limit INTEGER The training time limit (seconds). Defaults
to letting autogluon run until it's had
enough.
--help Show this message and exit.
Microsoft Visual C++ Redistributable is required, if you are on Windows.
autogluon uses the CPU version of PyTorch, by default. If you have a CUDA-enabled GPU, installing the CUDA version of PyTorch will greatly increase training speed:
pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116