Offline Handwritten Signature Verification Using a Stream-Based Approach

This repository provides tools for generating dissimilarity data, training and testing writer-independent classifiers in batch mode, and converting batch data into a stream of handwritten signatures. Additionally, it implements a prequential evaluation (test-then-train) approach to assess the performance of classifiers on the stream.

Installation

This package requires python 3. Installation can be done with pip:

pip install git+https://github.com/kdMoura/stream_hsv.git

You can also clone this repository and install it with pip install -e <path/to/repository>

Usage

Data generation

The data.py script is designed to generate dissimilarity data for both training and testing. Before applying the generation process, the datasets should:

1. Undergo preprocessing as outlined in https://github.com/tallesbrito/contrastive_sigver.
2. Have features extracted using the SigNet-S available in the same repository.
3. The resulting data should be a single .NPZ file containing:
   • features: shape (samples, features)
   • y: writer ID
   • yforg: 1 for forgery, 0 otherwise

To generate dissimilarity training datasets, use the training_generation subcommand:

python -m shsv.data training_generation --n-data 5  \
--input-data <data.npz> --f-output-path /path/to/output/folder  \
--n-gen 12 --include-users 300 350

To generate dissimilarity test datasets, use the test_generation subcommand:

python -m shsv.data test_generation --n-data 5  \
--input-data <data.npz> --f-output-path /path/to/output/folder \
--n-ref 12 --n-query 10 --include-users 0 300

Model Learning and Testing

The model.py script is used for batch mode training and testing and also for stream prequential (test-then-train) process.

1. Batch mode training:

python -m shsv.model batch_train --model sgd \
--f-train-path /path/to/train/datasets/folder --f-output-path /path/to/output/folder

2. Batch mode test:

python -m shsv.model batch_test --f-model-path /path/to/models/folder \
--f-test-path /path/to/test/data/folder  \
--f-output-path /path/to/output/folder

3. Stream prequential:

python -m shsv.model prequential --f-model-path /path/to/models/folder \
--f-test-path /path/to/test/data/folder \
--f-output-path /path/to/output/folder \
--n-ref 12 --chunk-size 600 --file-number 0

Evaluation

The evaluation.py script contains functions to compute metrics on batch and stream predictions files.

1. Batch evaluation:

python -m shsv.evaluation batch --f-input-path /path/to/predictions/folder --f-output-path /path/to/output/folder \
--n-ref 1 2 3 5 10 12 --forgery skilled --thr-type global --fusions max mean min median

2. Stream evaluation:

python -m shsv.evaluation stream --f-input-path /path/to/predictions/folder --f-output-path /path/to/output/folder \
--forgery skilled --thr-type global --fusions max \
--window-size 400 --window-step 400

Reproducibility Scripts

This repository includes three key scripts designed for the reproducibility of the experiments:

1. reproduce_data.py: This script includes all necessary configurations to generate the datasets used in the experiments.

2. reproduce_model.py: This script includes all configurations for training the models and performing tests on the datasets.

3. reproduce_evaluation.py: This script performs the evaluation of the model results, including all necessary configurations to replicate the evaluation metrics presented in the study.

Important Note:

While these scripts are fully configured and can be run as-is, it is not advisable to execute them sequentially due to the significant amount of time this would take. To optimize your workflow, it is recommended to adapt the execution by parallelizing the processes.

Citation

If you use our code, please consider citing it as:

de Moura, K.G., Cruz, R.M.O., Sabourin, R. (2025). "Offline Handwritten Signature Verification Using a Stream-Based Approach". In: Antonacopoulos, A., Chaudhuri, S., Chellappa, R., Liu, CL., Bhattacharya, S., Pal, U. (eds) Pattern Recognition. ICPR 2024. Lecture Notes in Computer Science, vol 15331. Springer, Cham. https://doi.org/10.1007/978-3-031-78119-3_19 (preprint)