---
license: apache-2.0
language:
- en
metrics:
- accuracy
base_model:
- liuhaotian/llava-v1.6-mistral-7b
pipeline_tag: image-classification
library_name: transformers
tags:
- llm
- mllm
- deepfake
---
# FFAA Model Card
## Model details
**Model type**: Face Forgery Analysis Assistant (FFAA) consists of a fine-tuned MLLM and Multi-answer Intelligent Decision System (MIDS).
It is a Multi-modal Large Language Model dedicated to the face forgery analysis.
Base MLLM: [liuhaotian/llava-v1.6-mistral-7b](https://huggingface.co/liuhaotian/llava-v1.6-mistral-7b)

**Paper or resources for more information**: [https://ffaa-vl.github.io/](https://ffaa-vl.github.io/)

**Where to send questions or comments about the model**: [https://github.com/thu-huangzc/FFAA/issues](https://github.com/thu-huangzc/FFAA/issues)

## Intended use
**Primary intended uses**: The primary use of FFAA is research on the applications of MLLMs in face forgery analysis, which is essential
for understanding the model’s decision-making process and advancing real-world face forgery analysis.

**Primary intended users**: The primary intended users of the model are researchers and hobbyists in computer vision, 
natural language processing, machine learning, and artificial intelligence.

## Training dataset
* 20K face forgery analysis VQA (FFA-VQA) dataset, captioned by GPT-4o.

* 90K historical answer data generated by the MLLM fine-tuned on FFA-VQA.

## Evaluation dataset
Open-World Face Forgery Analysis Benchmark (OW-FFA-Bench), including 6 face forgery generalization test sets. 
The download link is [Google driver](https://drive.google.com/file/d/1867ZKwFCh_OLm-uUsiIiI9RjrUv0JMZX/view?usp=drive_link)