Skip to content

Latest commit

 

History

History
47 lines (38 loc) · 1.95 KB

DATASET.md

File metadata and controls

47 lines (38 loc) · 1.95 KB
Raw

Dataset

To run the training and testing code, we require the following data organization format

${ROOT}--
        |--${DATASET1}
        |--${DATASET2}
        ...

The ROOT folder can be set in libs/dataset/data.py by setting ROOT = "path/to/root". Each sub-directory ${DATASET} should be the name of one specific dataset (e.g. DAVIS17 or Youtube-VOS) and contain all video and annotation data.

DAVIS Organization

To run the training script on davis16/17 dataset, please ensure the data is organized as following format

DAVIS16(DAVIS17)
      |----JPEGImages
      |----Annotations
      |----data
      |------|-----db_info.yaml

Where JPEGImages and Annotations contain the 480p frames and annotation masks of each video. The db_info.yaml contains the meta information of each video sequences and can be found at the davis evaluation repository.

Youtbe-VOS Organization

To run the training script on youtube-vos dataset, please ensure the data is organized as following format

Youtube-VOS
      |----train
      |     |-----JPEGImages
      |     |-----Annotations
      |     |-----meta.json
      |----valid
      |     |-----JPEGImages
      |     |-----Annotations
      |     |-----meta.json

Where JPEGImages and Annotations contain the frames and annotation masks of each video.

Custom Dataset

To run the codebase on custom dataset, please see the template and implement related interface and alias.

DALI Backend Support

Nvidia-dali backend loader is implemented for training on davis and youtube-vos. If you want to use the dali loader for efficient frame decode and transformation, you can install DALI first and set data_backend: 'DALI' in config.yaml or data_backend DALI in command line.

Note: DALI backend is only for training now