# License Plate Enhancement - From TV shows to reality
[](LICENSE)
Author: Zixuan Zhang, Chengxuan Cai
## Welcom to LicenseEnhancer
In Hollywood crimes movies we often see detectives solving crimes with the help from one of
their computer geeks who can reveal hidden information from blurred, low-quality images. This
project is an effort to achieve the same task, but on one specific type of image - license plates.
License plate enhancement is a detailed application of a broader field called *Single Image Super Resolution*
(SISR).
The project is inspired by several state-of-the-art SRSR models such as:
* [Photo-realistic single image super resolution using a Generative Adversarial Network](https://arxiv.org/abs/1609.04802)
* [Residual Dense Network for Image Super Resolution](https://arxiv.org/abs/1802.08797)
* [ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks](https://arxiv.org/abs/1809.00219)
* [Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network](https://arxiv.org/abs/1609.05158)
The dataset used in this project is called the [Chinese City Parking Dataset](https://github.com/detectRecog/CCPD), a large-scale collection of plate images in
various conditions.
Read my post on [Medium](https://towardsdatascience.com/license-plate-image-enhancement-5a170475bec1) for further understanding
## Gallery
## Requirement
Preprocessing
* Dask >= 2.11.0
* PIL >= 6.2.2
Training & Evaluation
* tensorflow >= 2.1.0
* numpy >= 1.18.1
* matplotlib >= 3.1.3
## Pipeline
Before training the model it is important to preprocess the raw dataset using the preprocess.py script
## Model Architecture
Our plate enhancer model is trained in an adversarial fashion(GAN), meaning the generator is trained to create realistic
reconstruction of images that can fool the discriminator, which is a binary classifier. Why GANs? Well, according to several
papers, GAN network tend to create more realistic image reconstruction comparing to model solely trained in the supervised
fashion. For instance, models that minimize Mean Square Error tend to have over-smoothing artifacts.
Therefore, there are two models - the generator(reconstructor) and the discriminator(classifier).
#### Generator
The generator is trained to minimize a novel hybrid loss function, namely the perceptual loss defined in the SRGAN paper
#### Discriminator
## Acknowledgement
I'd like to thank Olaoluwa Adigun for his amazing suggestions during the span of this project!
This project won the *Best Deep Learning Design Award* in USC EE599-Deep Learning. Here's the [link](https://drive.google.com/file/d/1VdCU_LArUe-KHj5vDVT0xyne9-bVK7n4/view)
to our amazing rojects done by my classmates!
Also, this project stands on the shoulder of many other SISR projects:
* [ESPCN](https://github.com/leftthomas/ESPCN)
* [SRGAN](https://github.com/tensorlayer/srgan)
* [ESRGAN](https://github.com/xinntao/ESRGAN)
人工智能项目资料-基于CNN多标签分类的车牌识别；renet残差网络.zip
版权申诉
17 浏览量
2024-02-07
15:35:46
上传
评论
收藏 25.72MB ZIP 举报
图像识别算法：基于CNN多标签分类的车牌识别；renet残差网络.zip （88个子文件） 
资料总结 picture 
pic.jpg 10KB plate_detect_and_recognition recog 
__init__.py 0B generate plate.py 2KB preprocess plate_enhange Enhance_Image.py 1KB LICENSE 1KB Utilities __init__.py 0B preprocess.py 5KB painter.py 982B trainVal.py 8KB lossMetric.py 2KB io.py 1KB Pretrained checkpoint 65B rrdb.data-00001-of-00002 27.34MB rrdb.index 7KB rrdb.data-00000-of-00002 16KB Tutorial_2_Training.ipynb 23KB .gitignore 2KB README.md 3KB Models __init__.py 0B SRResNet.py 3KB GAN.py 2KB RRDBNet.py 3KB ESPCN.py 1KB Training.py 3KB Tutorial_1_Enhance_Image.ipynb 360KB predict tensor __init__.py 0B cnn.py 1KB __init__.py 0B model tensor __init__.py 0B cnn.py 4KB __init__.py 0B layer tensor __init__.py 0B inception10.py 3KB vgg16.py 4KB alexnet8.py 2KB cnn.py 16KB lenet5.py 953B resnet18.py 3KB baseline.py 1KB __init__.py 0B main.py 2KB detect detect.py 6KB .gitignore 77B README.md 484B ocr __init__.py 0B generate plate.py 2KB tests tensor __init__.py 0B plate.py 2KB cnn.py 375B __init__.py 0B predict tensor __init__.py 0B cnn_plate.py 2KB resnet1.py 1KB cnn.py 1KB resnet.py 1KB __pycache__ resnet.cpython-37.pyc 2KB cnn.cpython-37.pyc 2KB cnn_plate.cpython-37.pyc 2KB __init__.cpython-37.pyc 142B resnet1.cpython-37.pyc 2KB __init__.py 0B __pycache__ __init__.cpython-37.pyc 135B model tensor __init__.py 0B cnn_plate.py 2KB resnet1.py 5KB cnn.py 4KB resnet.py 3KB __pycache__ resnet.cpython-37.pyc 3KB cnn.cpython-37.pyc 3KB cnn_plate.cpython-37.pyc 2KB __init__.cpython-37.pyc 140B resnet1.cpython-37.pyc 3KB base.py 4KB __init__.py 0B __pycache__ __init__.cpython-37.pyc 133B __pycache__ __init__.cpython-37.pyc 127B layer tensor __init__.py 0B cnn_plate.py 1KB resnet1.py 3KB cnn.py 1KB resnet.py 3KB __pycache__ resnet.cpython-37.pyc 2KB cnn.cpython-37.pyc 1KB cnn_plate.cpython-37.pyc 1KB __init__.cpython-37.pyc 140B resnet1.cpython-37.pyc 3KB __init__.py 0B __pycache__ __init__.cpython-37.pyc 133B资源评论
