README.md

---
size_categories:
- 1M<n<10M
---
# PreFM: Online Audio-Visual Event Parsing via Predictive Future Modeling
This is the dataset of the NeurIPS 2025 paper "[PreFM: Online Audio-Visual Event Parsing via Predictive Future Modeling](https://arxiv.org/abs/2505.23155)".

The corresponding github repo is [PreFM](https://github.com/XiaoYu-1123/PreFM).

**Directory Structure:**
    Please organize the downloaded datasets into the following structure under the `dataset` directory:

    ```
    dataset
    |-- LLP
    |   |-- data
    |   |   |-- CLAP
    |   |   |-- CLIP
    |   |   `-- st
    |   `-- label
    |       |-- feature_label   # Labels for features
    |       |   |-- audio
    |       |   `-- visual
    |       `-- label           # Event category labels
    |           |-- audio
    |           `-- visual
    |-- UnAV-100
    |   |-- data
    |   |   |-- CLAP
    |   |   |-- CLIP
    |   |   `-- onepeace        # Audio-visual data processed by OnePeace
    |   `-- label
    |       |-- feature_label   # Labels for features
    |       |   |-- test
    |       |   |-- train
    |       |   `-- val
    |       |-- feature_label_modified # Feature labels for PreFM+ on the On AVEL task
    |       |   |-- test
    |       |   |-- train
    |       |   `-- val
    |       |-- label_clip_clap # Event category labels
    |       |   |-- test
    |       |   |-- train
    |       |   `-- val
    |       `-- label_onepeace  # Event category labels for PreFM+ on the On AVEL task
    |           |-- test
    |           |-- train
    |           `-- val
    ```