This directory contains the more detailed documentation that sits underneath the top-level README.md.
Start with the top-level README if you are new to the project. It explains:
- the config tree
- the CLI style
- run directory layout
- logger selection
- the extension path for datasets and models
Then use the topic-specific docs below.
- ../README.md: main project overview, CLI, config model, extension points
- datasets/DATASET_STRUCTURE.md: current split-file expectations and dataset layout rules
- augmentation/policies.md: augmentation policy structure, available shared policies, and extension path
- precomputed_features.md: HDF5 feature training and extraction
- mil.md: MIL and CLAM over precomputed bags
- video.md: video encoders, intermediates, and framewise heads
- testing/TEST_RUN_README.md: simple smoke-test examples
- ../tests/README.md: test suite structure and how to run it
If you want to:
- run a normal experiment:
- start with ../README.md
- understand dataset structure or add a dataset:
- change augmentation defaults or add new policies:
- train from extracted features:
- use CLAM / bag-of-features MIL:
- read mil.md
- use video encoders or framewise heads:
- read video.md