metadata
license: apache-2.0
language:
- en
pipeline_tag: image-to-text
library_name: transformers
base_model: Qwen/Qwen-VL
tags:
- vision-language
- chart-understanding
- chart-question-answering
- document-understanding
- multimodal
datasets:
- custom
metrics:
- accuracy
model-index:
- name: ChartQwen
results: []
ChartQwen
Model Description
ChartQwen is a vision-language model fine-tuned from Qwen/Qwen-VL for chart understanding tasks.
The model is designed to interpret visual charts such as bar charts, line graphs, and plots, and answer natural language questions related to them.
It supports multimodal reasoning by jointly processing images and text prompts.
Intended Use
This model can be used for:
- Chart question answering
- Chart data interpretation
- Visual reasoning over plots and graphs
- Document and report analysis involving charts
Training Details
- Base model: Qwen/Qwen-VL
- Modality: Image + Text
- Fine-tuning type: Supervised fine-tuning on chart-related visual-question pairs
- Dataset: Custom chart dataset (generated and curated for chart understanding)
Limitations
- Performance may degrade on low-resolution or highly cluttered charts
- The model may struggle with handwritten charts or uncommon chart styles
- Numerical precision depends on chart clarity