Fast & Lightweight License Plate OCR

---

Introduction

Lightweight and fast OCR models for license plate text recognition. You can train models from scratch or use the trained models for inference.

The idea is to use this after a plate object detector, since the OCR expects the cropped plates.

Features

• Keras 3 Backend Support: Train seamlessly using TensorFlow, JAX, or PyTorch backends 🧠
• Augmentation Variety: Diverse training-time augmentations via Albumentations library 🖼️
• Efficient Execution: Lightweight models that are cheap to run 💰
• ONNX Runtime Inference: Fast and optimized inference with ONNX runtime ⚡
• User-Friendly CLI: Simplified CLI for training and validating OCR models 🛠️
• Model HUB: Access to a collection of pre-trained models ready for inference 🌟
• Train/Fine-tune: Easily train or fine-tune your own models 🔧
• Export-Friendly: Export easily to CoreML or TFLite formats 📦

Available Models

Optimized, ready to use models with config files for inference or fine-tuning.

Model Name	Size	Arch	b=1 Avg. Latency (ms)	Plates/sec (PPS)	Model Config	Plate Config	Val Results
cct-s-v1-global-model	S	CCT	0.5877	1701.63	model_config.yaml	plate_config.yaml	results
cct-xs-v1-global-model	XS	CCT	0.3232	3094.21	model_config.yaml	plate_config.yaml	results

Tip

🚀 Try the above models in Hugging Spaces.

Note

Benchmark Setup

These results were obtained with:

• Hardware: NVIDIA RTX 3090 GPU
• Execution Providers: ['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']
• Install dependencies: pip install fast-plate-ocr[onnx-gpu]

Inference

For doing inference, install:

pip install fast-plate-ocr[onnx-gpu]

By default, no ONNX runtime is installed. To run inference, you must install at least one ONNX backend using an appropriate extra.

Platform/Use Case	Install Command	Notes
CPU (default)	pip install fast-plate-ocr[onnx]	Cross-platform
NVIDIA GPU (CUDA)	pip install fast-plate-ocr[onnx-gpu]	Linux/Windows
Intel (OpenVINO)	pip install fast-plate-ocr[onnx-openvino]	Best on Intel CPUs
Windows (DirectML)	pip install fast-plate-ocr[onnx-directml]	For DirectML support
Qualcomm (QNN)	pip install fast-plate-ocr[onnx-qnn]	Qualcomm chipsets

Usage

To predict from disk image:

from fast_plate_ocr import LicensePlateRecognizer

m = LicensePlateRecognizer('cct-xs-v1-global-model')
print(m.run('test_plate.png'))

Run demo

To run model benchmark:

from fast_plate_ocr import LicensePlateRecognizer

m = LicensePlateRecognizer('cct-xs-v1-global-model')
m.benchmark()

Benchmark demo

Training

You can train models from scratch or fine-tune a pre-trained one using your own license plate dataset.

Install the training dependencies:

pip install fast-plate-ocr[train]

Fine-tuning Tutorial

A complete tutorial notebook is available for fine-tuning a license plate OCR model on your own dataset: examples/fine_tune_workflow.ipynb. It covers the full workflow, from preparing your dataset to training and exporting the model.

For full details on data preparation, model configs, fine-tuning, and training commands, check out the docs.

Contributing

Contributions to the repo are greatly appreciated. Whether it's bug fixes, feature enhancements, or new models, your contributions are warmly welcomed.

To start contributing or to begin development, you can follow these steps:

1. Clone repo

   git clone https://github.com/ankandrew/fast-plate-ocr.git

2. Install all dependencies (make sure you have Poetry installed):

   make install

3. To ensure your changes pass linting and tests before submitting a PR:

   make checks

Citations

@article{hassani2021escaping,
    title   = {Escaping the Big Data Paradigm with Compact Transformers},
    author  = {Ali Hassani and Steven Walton and Nikhil Shah and Abulikemu Abuduweili and Jiachen Li and Humphrey Shi},
    year    = 2021,
    url     = {https://arxiv.org/abs/2104.05704},
    eprint  = {2104.05704},
    archiveprefix = {arXiv},
    primaryclass = {cs.CV}
}