VBench_Leaderboard / constants.py
ynhe's picture
Update constants.py
a635b9d verified
raw
history blame
6.36 kB
import os
# this is .py for store constants
MODEL_INFO = [
"Model Name (clickable)",
"Total Score",
"Quality Score",
"Semantic Score",
"Selected Score",
]
TASK_INFO = [
"subject consistency",
"background consistency",
"temporal flickering",
"motion smoothness",
"aesthetic quality",
"imaging quality",
"dynamic degree",
"object class",
"multiple objects",
"human action",
"color",
"spatial relationship",
"scene",
"appearance style",
"temporal style",
"overall consistency"]
DEFAULT_INFO = [
"subject consistency",
"background consistency",
"temporal flickering",
"motion smoothness",
"aesthetic quality",
"imaging quality",
]
QUALITY_LIST = [
"subject consistency",
"background consistency",
"temporal flickering",
"motion smoothness",
"aesthetic quality",
"imaging quality",
"dynamic degree",]
SEMANTIC_LIST = [
"object class",
"multiple objects",
"human action",
"color",
"spatial relationship",
"scene",
"appearance style",
"temporal style",
"overall consistency"
]
DIM_WEIGHT = {
"subject consistency":1,
"background consistency":1,
"temporal flickering":1,
"motion smoothness":1,
"aesthetic quality":1,
"imaging quality":1,
"dynamic degree":0.5,
"object class":1,
"multiple objects":1,
"human action":1,
"color":1,
"spatial relationship":1,
"scene":1,
"appearance style":1,
"temporal style":1,
"overall consistency":1
}
SEMANTIC_WEIGHT = 1
QUALITY_WEIGHT = 4
DATA_TITILE_TYPE = ['markdown', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number', 'number']
SUBMISSION_NAME = "vbench_leaderboard_submission"
SUBMISSION_URL = os.path.join("https://huggingface.co/datasets/Vchitect/", SUBMISSION_NAME)
CSV_DIR = "./vbench_leaderboard_submission/results.csv"
COLUMN_NAMES = MODEL_INFO + TASK_INFO
LEADERBORAD_INTRODUCTION = """# VBench Leaderboard
*"Which Video Generation Model is better?"*
πŸ† Welcome to the leaderboard of the **VBench**! 🎦 *A Comprehensive Benchmark Suite for Video Generative Models*
<div style="display: flex; flex-wrap: wrap; align-items: center; gap: 10px;">
<a href='https://arxiv.org/abs/2311.17982'><img src='https://img.shields.io/badge/cs.CV-Paper-b31b1b?logo=arxiv&logoColor=red'></a>
<a href='https://vchitect.github.io/VBench-project/'><img src='https://img.shields.io/badge/VBench-Website-green?logo=googlechrome&logoColor=green'></a>
<a href='https://pypi.org/project/vbench/'><img src='https://img.shields.io/pypi/v/vbench'></a>
<a href='https://www.youtube.com/watch?v=7IhCC8Qqn8Y'><img src='https://img.shields.io/badge/YouTube-Video-c4302b?logo=youtube&logoColor=red'></a>
<a href='https://hits.seeyoufarm.com'><img src='https://hits.seeyoufarm.com/api/count/incr/badge.svg?url=https%3A%2F%2Fgithub.com%2FVchitect%2FVBench&count_bg=%23FFA500&title_bg=%23555555&icon=&icon_color=%23E7E7E7&title=visitors&edge_flat=false'></a>
</div>
- **Comprehensive Dimensions:** We carefully decompose video generation quality into 16 comprehensive dimensions to reveal individual model's strengths and weaknesses.
- **Human Alignment:** We conducted extensive experiments and human annotations to validate robustness of VBench.
- **Valuable Insights:** VBench provides multi-perspective insights useful for the community.
Please follow the instructions in [VBench](https://github.com/Vchitect/VBench?tab=readme-ov-file#usage) to upload the generated `result.json` file here. After clicking the `Submit Eval` button, click the `Refresh` button.
"""
SUBMIT_INTRODUCTION = """# Submit on VBench Benchmark Introduction
## ⚠ Please note that you need to obtain the file `evaluation_results/*eval_results.json` by running [VBench Github](https:) and upload the evaluation results.
Uploading generated videos or images of the model is invalid!
"""
TABLE_INTRODUCTION = """
"""
LEADERBORAD_INFO = """
VBench, a comprehensive benchmark suite for video generative models. We design a comprehensive and hierarchical Evaluation Dimension Suite to decompose "video generation quality" into multiple well-defined dimensions to facilitate fine-grained and objective evaluation. For each dimension and each content category, we carefully design a Prompt Suite as test cases, and sample Generated Videos from a set of video generation models. For each evaluation dimension, we specifically design an Evaluation Method Suite, which uses carefully crafted method or designated pipeline for automatic objective evaluation. We also conduct Human Preference Annotation for the generated videos for each dimension, and show that VBench evaluation results are well aligned with human perceptions. VBench can provide valuable insights from multiple perspectives.
"""
CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
CITATION_BUTTON_TEXT = r"""@article{huang2023vbench,
title={{VBench}: Comprehensive Benchmark Suite for Video Generative Models},
author={Huang, Ziqi and He, Yinan and Yu, Jiashuo and Zhang, Fan and Si, Chenyang and Jiang, Yuming and Zhang, Yuanhan and Wu, Tianxing and Jin, Qingyang and Chanpaisit, Nattapol and Wang, Yaohui and Chen, Xinyuan and Wang, Limin and Lin, Dahua and Qiao, Yu and Liu, Ziwei},
journal={arXiv preprint arXiv:2311.17982},
year={2023}
}"""
NORMALIZE_DIC = {
"subject consistency": {"Min": 0.1462, "Max": 1.0},
"background consistency": {"Min": 0.2615, "Max": 1.0},
"temporal flickering": {"Min": 0.6293, "Max": 1.0},
"motion smoothness": {"Min": 0.706, "Max": 0.9975},
"dynamic degree": {"Min": 0.0, "Max": 1.0},
"aesthetic quality": {"Min": 0.0, "Max": 1.0},
"imaging quality": {"Min": 0.0, "Max": 1.0},
"object class": {"Min": 0.0, "Max": 1.0},
"multiple objects": {"Min": 0.0, "Max": 1.0},
"human action": {"Min": 0.0, "Max": 1.0},
"color": {"Min": 0.0, "Max": 1.0},
"spatial relationship": {"Min": 0.0, "Max": 1.0},
"scene": {"Min": 0.0, "Max": 0.8222},
"appearance style": {"Min": 0.0009, "Max": 0.2855},
"temporal style": {"Min": 0.0, "Max": 0.364},
"overall consistency": {"Min": 0.0, "Max": 0.364}
}