README.md

---
license: apache-2.0
datasets:
- TIGER-Lab/VideoFeedback
language:
- en
metrics:
- accuracy/spcc
library_name: transformers
pipeline_tag: visual-question-answering
---


[📃Paper] | [🌐Website](https://tiger-ai-lab.github.io/MantisScore/) | [💻Github](https://github.com/TIGER-AI-Lab/MantisScore) | [🛢️Datasets](https://huggingface.co/datasets/TIGER-Lab/VideoFeedback) | [🤗Model](https://huggingface.co/TIGER-Lab/MantisScore) | [🤗Model-variant](https://huggingface.co/TIGER-Lab/MantisScore-anno-only) | [🤗Demo](https://huggingface.co/spaces/Mantis-VL/MantisScore)


![MantisScore](https://tiger-ai-lab.github.io/MantisScore/static/images/teaser.png)

## Introduction
- MantisScore is a video quality evaluation model, taking [Mantis-8B-Idefics2](https://huggingface.co/TIGER-Lab/Mantis-8B-Idefics2) as base-model
and trained on [VideoFeedback](https://huggingface.co/datasets/TIGER-Lab/VideoFeedback),
a large video evaluation dataset with multi-aspect human scores. 

- MantisScore can reach 75+ Spearman correlation with humans on VideoEval-test, surpassing all the MLLM-prompting methods and feature-based metrics. 

- MantisScore also beat the best baselines on other three benchmarks EvalCrafter, GenAI-Bench and VBench, showing high alignment with human evaluations.

## Evaluation Results

We test our video evaluation model MantisScore on VideoEval-test, EvalCrafter, GenAI-Bench and VBench.
For the first two benchmarks, we take Spearman corrleation between model's output and human ratings 
averaged among all the evaluation aspects as indicator. 
For GenAI-Bench and VBench, which include human preference data among two or more videos, 
we employ the model's output to predict preferences and use pairwise accuracy as the performance indicator.

Moreover, we use [MantisScore](https://huggingface.co/TIGER-Lab/MantisScore) trained on VideoFeedback dataset 
for VideoFeedback-test set, while for other three benchmarks, we use 
[MantisScore-anno-only](https://huggingface.co/TIGER-Lab/MantisScore-anno-only) variant trained on VideoFeedback dataset 
with real videos excluded. 

The evaluation results are shown below: 


| metric            | Final Avg Score | VideoFeedback-test | EvalCrafter | GenAI-Bench | VBench     |
|:-----------------:|:--------------:|:--------------:|:-----------:|:-----------:|:----------:|
| MantisScore (reg) |       **69.6** |           75.7 |    **51.1** |    **78.5** |   **73.0** |
| MantisScore (gen) |           55.6 |       **77.1** |        27.6 |        59.0 |       58.7 |
| Gemini-1.5-Pro    |    <u>39.7</u> |           22.1 |        22.9 |        60.9 |       52.9 |
| Gemini-1.5-Flash  |           39.4 |           20.8 |        17.3 | <u>67.1</u> |       52.3 |
| GPT-4o            |           38.9 |    <u>23.1</u> |        28.7 |        52.0 |       51.7 |
| CLIP-sim          |           31.7 |            8.9 | <u>36.2</u> |        34.2 |       47.4 |
| DINO-sim          |           30.3 |            7.5 |        32.1 |        38.5 |       43.3 |
| SSIM-sim          |           29.5 |           13.4 |        26.9 |        34.1 |       43.5 |
| CLIP-Score        |           28.6 |           -7.2 |        21.7 |        45.0 |       54.9 |
| LLaVA-1.5-7B      |           27.1 |            8.5 |        10.5 |        49.9 |       39.4 |
| LLaVA-1.6-7B      |           23.3 |           -3.1 |        13.2 |        44.5 |       38.7 |
| X-CLIP-Score      |           23.2 |           -1.9 |        13.3 |        41.4 |       40.1 |
| PIQE              |           19.6 |          -10.1 |        -1.2 |        34.5 |<u> 55.1</u>|
| BRISQUE           |           19.0 |          -20.3 |         3.9 |        38.5 |       53.7 |
| Idefics2          |           18.3 |            6.5 |         0.3 |        34.6 |       31.7 |
| MSE-dyn           |           10.6 |           -5.5 |       -17.0 |        28.4 |       36.5 |
| SSIM-dyn          |            9.2 |          -12.9 |       -26.4 |        31.4 |       44.5 |
<!-- | Fuyu              |               - |              - |           - |           - |          - |
| Kosmos-2          |               - |              - |           - |           - |          - |
| CogVLM            |               - |              - |           - |           - |          - |
| OpenFlamingo      |               - |              - |           - |           - |          - | -->

The best in MantisScore series is in bold and the best in baselines is underlined. 
<!-- "-" means the answer of MLLM is meaningless or in wrong format. -->

## Usage
### Installation
```bash
pip install git+https://github.com/TIGER-AI-Lab/MantisScore.git
```

### Inference
```shell
cd examples
```

```python
import av
import numpy as np
from typing import List
import torch
from transformers import AutoProcessor
from models.idefics2 import Idefics2ForSequenceClassification

def _read_video_pyav(
    frame_paths:List[str], 
    max_frames:int,
):
    frames = []
    container.seek(0)
    start_index = indices[0]
    end_index = indices[-1]
    for i, frame in enumerate(container.decode(video=0)):
        if i > end_index:
            break
        if i >= start_index and i in indices:
            frames.append(frame)
    return np.stack([x.to_ndarray(format="rgb24") for x in frames])

MAX_NUM_FRAMES=16
ROUND_DIGIT=4
REGRESSION_QUERY_PROMPT = """
Suppose you are an expert in judging and evaluating the quality of AI-generated videos,
please watch the following frames of a given video and see the text prompt for generating the video,
then give scores from 5 different dimensions:
(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color
(2) temporal consistency, both the consistency of objects or humans and the smoothness of motion or movements
(3) dynamic degree, the degree of dynamic changes
(4) text-to-video alignment, the alignment between the text prompt and the video content
(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge

for each dimension, output a float number from 1.0 to 4.0,
the higher the number is, the better the video performs in that sub-score, 
the lowest 1.0 means Bad, the highest 4.0 means Perfect/Real (the video is like a real video)
Here is an output example:
visual quality: 3.2
temporal consistency: 2.7
dynamic degree: 4.0
text-to-video alignment: 2.3
factual consistency: 1.8

For this video, the text prompt is "{text_prompt}",
all the frames of video are as follows:
"""

video_path="examples/video1.mp4"
video_prompt="Near the Elephant Gate village, they approach the haunted house at night. Rajiv feels anxious, but Bhavesh encourages him. As they reach the house, a mysterious sound in the air adds to the suspense."

processor = AutoProcessor.from_pretrained(f"TIGER-Lab/MantisScore",torch_dtype=torch.bfloat16)
model = Idefics2ForSequenceClassification.from_pretrained(f"TIGER-Lab/MantisScore",torch_dtype=torch.bfloat16).eval()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)

# sample uniformly 8 frames from the video
container = av.open(video_path)
total_frames = container.streams.video[0].frames
if total_frames > MAX_NUM_FRAMES:
    indices = np.arange(0, total_frames, total_frames / MAX_NUM_FRAMES).astype(int)
else:
    indices = np.arange(total_frames)

frames = [Image.fromarray(x) for x in _read_video_pyav(container, indices)]
eval_prompt = REGRESSION_QUERY_PROMPT.format(text_prompt=video_prompt)
num_image_token = eval_prompt.count("<image>")
if num_image_token < len(frames):
    eval_prompt += "<image> " * (len(frames) - num_image_token)

flatten_images = []
for x in [frames]:
    if isinstance(x, list):
        flatten_images.extend(x)
    else:
        flatten_images.append(x)
flatten_images = [Image.open(x) if isinstance(x, str) else x for x in flatten_images]
inputs = processor(text=eval_prompt, images=flatten_images, return_tensors="pt")
inputs = {k: v.to(model.device) for k, v in inputs.items()}

with torch.no_grad():
    outputs = model(**inputs)

logits = outputs.logits
num_aspects = logits.shape[-1]

aspect_scores = []
for i in range(num_aspects):
    aspect_scores.append(round(logits[0, i].item(),ROUND_DIGIT))
print(aspect_scores)

"""
# model output on visual quality, temporal consistency, dynamic degree,
# text-to-video alignment, factual consistency, respectively
[2.2969, 2.4375, 2.8281, 2.5, 2.4688]
"""

```

### Training
see [MantisScore/training](https://github.com/TIGER-AI-Lab/MantisScore/training) for details

### Evaluation
see [MantisScore/benchmark]((https://github.com/TIGER-AI-Lab/MantisScore/benchmark)) for details

## Citation