MomentSeeker_1k8 / README.md
memray's picture
Update README.md
968b24a verified
|
Raw History Blame Contribute Delete
1.09 kB
metadata
dataset_info:
  features:
    - name: query
      dtype: string
    - name: negative_frames
      list:
        - name: start_time
          dtype: float64
        - name: end_time
          dtype: float64
        - name: output_path
          dtype: string
        - name: frame_path
          dtype: string
    - name: positive_frames
      list:
        - name: start_time
          dtype: float64
        - name: end_time
          dtype: float64
        - name: output_path
          dtype: string
        - name: frame_path
          dtype: string
    - name: input_frames
      dtype: string
  splits:
    - name: test
      num_bytes: 1984070
      num_examples: 1800
  download_size: 380078
  dataset_size: 1984070
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*

Original MomentSeeker data (jsonl file). It contains 1,800 rows and duplicate queries (discussion).

Check out https://huggingface.co/datasets/VLM2Vec/MomentSeeker for the dedup version.