---
license: apache-2.0
extra_gated_prompt: >-
  You agree to not use the dataset to conduct experiments that cause harm to
  human subjects. Please note that the data in this dataset may be subject to
  other agreements. Before using the data, be sure to read the relevant
  agreements carefully to ensure compliant use. Video copyrights belong to the
  original video creators or platforms and are for academic research use only.
dataset_info:
  config_name: dream-1k
  features:
  - name: idx
    dtype: string
  - name: video_file
    dtype: string
  - name: source
    dtype: string
  - name: duration
    dtype: string
  - name: description
    dtype: string
  - name: events
    dtype: string
  - name: n_subjects
    dtype: string
  - name: n_shots
    dtype: string
  - name: n_events
    dtype: string
  splits:
  - name: test
size_categories:
- 1K<n<10K
configs:
- config_name: DREAM-1K
  data_files:
  - split: test
    path: json/metadata.json
---
# DREAM-1K

## Dataset Description

- **Repository:** [DREAM-1K](https://github.com/bytedance/tarsier)
- **Paper:** [2407.00634](https://arxiv.org/abs/2407.00634)

DREAM-1K (<b>D</b>escription
with <b>R</b>ich <b>E</b>vents, <b>A</b>ctions, and <b>M</b>otions) is a challenging video description benchmark. It contains a collection of 1,000 short (around 10 seconds) video clips with diverse complexities from five different origins: live-action movies, animated movies, stock videos, long YouTube videos, and TikTok-style short videos. We provide a fine-grained manual annotation for each video.

Bellow is the dataset statistics:

![images](./assets/dream-1k-statistics.png)