Mels22 commited on
Commit
a67ab2d
·
verified ·
1 Parent(s): 7e63a5e

Update the README file

Browse files
Files changed (1) hide show
  1. README.md +70 -0
README.md CHANGED
@@ -28,4 +28,74 @@ configs:
28
  path: data/train-*
29
  - split: test
30
  path: data/test-*
 
 
 
 
 
 
 
 
 
 
31
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
28
  path: data/train-*
29
  - split: test
30
  path: data/test-*
31
+ license: apache-2.0
32
+ task_categories:
33
+ - image-classification
34
+ - object-detection
35
+ tags:
36
+ - signature
37
+ - document
38
+ pretty_name: Signature Detection and Verification
39
+ size_categories:
40
+ - 10K<n<100K
41
  ---
42
+
43
+ # Signature Detection and Verification Dataset
44
+
45
+ A comprehensive dataset designed for building and evaluating **end-to-end signature analysis pipelines**, including **signature detection** in document images and **signature verification** using genuine/forged pair classification.
46
+
47
+ **Developed by**:[@Mels22](https://huggingface.co/Mels22) and [@JoeCao](https://huggingface.co/JoeCao)
48
+
49
+ ## Pipeline Overview
50
+
51
+ This dataset supports a complete **signature detection and verification pipeline**. The process involves identifying the signature in a document and comparing it with a reference to determine if it is genuine or forged.
52
+
53
+ ![Detection and Verification Pipeline](pipeline.png)
54
+
55
+ - The **Detection Model** locates the signature in the document.
56
+ - The cropped signature (`to_verify_signature`) is passed along with a sample signature (`sample_signature`) to the **Verification Model**.
57
+ - The model then classifies the signature as either Genuine or Forged.
58
+
59
+
60
+ ## Dataset Summary
61
+
62
+ | Split | Samples |
63
+ |-------|---------|
64
+ | Train | 23,206 |
65
+ | Test | 6,195 |
66
+ | **Total** | **29,401** |
67
+
68
+ This dataset supports two key tasks:
69
+ - **Detection:** Identifying the bounding boxes of signatures in scanned document images.
70
+ - **Verification:** Comparing a signature within the document to a reference (sample) signature to determine whether it's **genuine** (`label = 0`) or **forged** (`label = 1`).
71
+
72
+ ## Features
73
+
74
+ Each sample in the dataset contains the following fields:
75
+
76
+ - `document` *(Image)*: The full document image that contains one or more handwritten signatures.
77
+ - `bbox` *(List of Bounding Boxes)*: The coordinates of the signature(s) detected in the `document`. Format: `[x_min, y_min, x_max, y_max]`.
78
+ - `to_verify_signature` *(Image)*: A cropped signature from the document image that needs to be verified.
79
+ - `sample_signature` *(Image)*: A standard reference signature used for comparison.
80
+ - `label` *(int)*: Indicates if the `to_verify_signature` is **genuine (0)** or **forged (1)** when compared to the `sample_signature`.
81
+
82
+ ## Data Sources & Construction
83
+
84
+ This dataset is **constructed by combining and modifying two publicly available datasets**:
85
+
86
+ - **Signature Images** were sourced from the [Kaggle Signature Verification Dataset](https://www.kaggle.com/datasets/robinreni/signature-verification-dataset), which provides genuine and forged signatures from multiple individuals for verification tasks.
87
+
88
+ - **Document Images with Signature Bounding Boxes** were taken from the [Signature Detection Dataset by NanoNets](https://github.com/NanoNets/SignatureDetectionDataset), which contains scanned documents with annotated signature regions.
89
+
90
+ ### How This Dataset Was Created
91
+
92
+ To create a seamless, unified pipeline dataset for **detection + verification**, the following modifications were made:
93
+
94
+ - **Synthetic Placement**: Signature images were programmatically inserted into real documents at their correct signing regions (e.g., bottom of the page or designated signature lines).
95
+ - **Blending with Background**: Signatures were rendered with varying opacities, filters, and transformations to match the document background, mimicking real-world signature scans.
96
+ - **Labeling and BBoxes**: The new locations of the inserted signatures were used to generate accurate bounding boxes for detection tasks.
97
+ - **Pairing for Verification**: Each inserted signature (`to_verify_signature`) was paired with a reference (`sample_signature`) and assigned a label: `0` for genuine or `1` for forged.
98
+
99
+ This process enables researchers to train and evaluate models for **both signature localization and signature verification** in a realistic, document-centric setting.
100
+
101
+