简介 Intro
这是一种多位数门牌号识别模型，采用了基于 PyTorch 框架的深度卷积神经网络实现方案，旨在从街景图像中高效而准确地识别多位数门牌号。模型的训练数据来源于谷歌街景门牌号码数据集（SVHN），每个图像中包含一组 0-9 的阿拉伯数字。通过广泛的测试，该模型在精确度方面取得了显著成绩，达到了89%。深度卷积神经网络结构的运用使得模型能够有效地捕捉门牌号图像中的数字特征，从而实现了对多位数门牌号的精准识别，在数字识别领域为街景应用提供了可靠的技术支持。
This is a model for recognizing multi-digit house numbers, employing a deep convolutional neural network based on the PyTorch framework, designed to identify multi-digit house numbers from street view images with high efficiency and accuracy. The training data for the model comes from the Google Street View House Numbers dataset (SVHN), which includes a series of Arabic numerals from 0 to 9 in each image. The model has achieved significant results in precision, with an accuracy rate of 89% after rigorous testing. The deep convolutional neural network structure allows the model to effectively capture the characteristics of the numbers in the house number images, thus enabling accurate recognition of multi-digit house numbers. It offers dependable technical support for street view applications in the realm of digital recognition.
在线体验 Demo
https://www.modelscope.cn/studios/MuGemSt/svhn
维护 Maintenance
git clone [email protected]:MuGemSt/svhn
cd svhn
使用 Usage
from modelscope import snapshot_download
model_dir = snapshot_download("MuGemSt/svhn")
数据集 Dataset
https://www.modelscope.cn/datasets/MuGemSt/svhn
训练曲线 Training curve
镜像 Mirror
https://www.modelscope.cn/models/MuGemSt/svhn
参考引用 Reference
[1] https://github.com/MuGemSt/SVHN-Recognition
[2] Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks