README.md

---
license: cc-by-4.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
dataset_info:
  features:
  - name: image_name
    dtype: string
  - name: image
    dtype: image
  - name: width
    dtype: int32
  - name: height
    dtype: int32
  - name: instances
    sequence:
    - name: category_id
      dtype: int32
    - name: mask
      sequence:
        sequence: float32
  splits:
  - name: train
    num_bytes: 8566253.0
    num_examples: 200
  - name: val
    num_bytes: 4532786.0
    num_examples: 100
  - name: test
    num_bytes: 3809329.0
    num_examples: 100
  download_size: 16498520
  dataset_size: 16908368.0
---

<h1 align="center">Line Graphics Digitization: A Step Towards Full Automation</h1>

<p align="center">
  <em>Omar Moured, Jiaming Zhang, Alina Roitberg, Thorsten Schwarz, Rainer Stiefelhagen</em>
</p>

<p align="center">
  <a href="https://link.springer.com/chapter/10.1007/978-3-031-41734-4_27">
    <img src="https://img.shields.io/badge/Paper-Springer-blue?logo=book" alt="Paper">
  </a><a href="https://huggingface.co/datasets/omoured/line-graphics-dataset/tree/main">
    <img src="https://img.shields.io/badge/Dataset-HuggingFace-orange?logo=huggingface" alt="Dataset">
  </a>
</p>

![Example visualization](./lgsample.png)

## Dataset Summary
The dataset includes instance segmentation masks for **400 real line chart images, manually labeled into 11 categories** by professionals. 
These images were collected from 5 different professions to enhance diversity. In our paper, we studied two levels of segmentation: **coarse-level**,
where we segmented (spines, axis-labels, legend, lines, titles), and **fine-level**, where we further segmented each category into x and y subclasses
(except for legend and lines), and individually segmented each line.

<!-- ## Category ID Reference
```python
class_id_mapping = {
    "Label": 0,
    "Legend": 1,
    "Line": 2,
    "Spine": 3,
    "Title": 4,
    "ptitle": 5,
    "xlabel": 6,
    "xspine": 7,
    "xtitle": 8,
    "ylabel": 9,
    "yspine": 10,
    "ytitle": 11
}
``` -->

## Dataset structure (train, validation, test)
- **image** - contains the PIL image of the chart
- **image_name** - image name with PNG extension
- **width** - original image width
- **height** - original image height
- **instances** - contains **N** number of COCO format instances. Check the sample visulization code below.

## Sample Usage
[optional] install `pycocotools` to rendeder masks with below code.

```python
from datasets import load_dataset
from pycocotools import mask
import matplotlib.pyplot as plt
import random

# Load dataset
ds = load_dataset("omoured/line-graphics-dataset")

# Class ID to name
id_to_name = {
    0: "Label", 1: "Legend", 2: "Line", 3: "Spine",
    4: "Title", 5: "ptitle", 6: "xlabel", 7: "xspine",
    8: "xtitle", 9: "ylabel", 10: "yspine", 11: "ytitle"
}

# Random image + instance
sample = random.choice(ds["val"])
img = sample["image"]
i = random.randint(0, len(sample["instances"]["mask"]) - 1)

# Get mask + class info
poly = sample["instances"]["mask"][i]
cat_id = sample["instances"]["category_id"][i]
cat_name = id_to_name.get(cat_id, "Unknown")

# Decode and plot
rle = mask.frPyObjects(poly, sample["height"], sample["width"])
bin_mask = mask.decode(rle)

plt.imshow(img)
plt.imshow(bin_mask, alpha=0.5, cmap="jet")
plt.title(f"imgname: {sample['image_name']}, inst: {cat_name}")
plt.axis("off")
plt.show()
```

## Copyrights
This dataset is published under the CC-BY 4.0 license, which allows for unrestricted usage, but it should be cited when used.

## Citation
```bibtex
@inproceedings{moured2023line,
  title={Line Graphics Digitization: A Step Towards Full Automation},
  author={Moured, Omar and Zhang, Jiaming and Roitberg, Alina and Schwarz, Thorsten and Stiefelhagen, Rainer},
  booktitle={International Conference on Document Analysis and Recognition},
  pages={438--453},
  year={2023},
  organization={Springer}
}
```

## Contact

If you have any questions or need further assistance with this dataset, please feel free to contact us:

- **Omar Moured**, omar.moured@kit.edu