Segmentation

Get started
DeepLab 是用于语义图像分割的最先进的深度学习模型,其目标是为图像中的每个像素分配语义标签(例如人,狗,猫)。
Download starter model
How it works
语义图像分割预测图像的每个像素是否与某个类相关联。这与检测矩形区域中目标[目标检测]的任务(https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/models/object_detection/overview.md)和对整个图像进行分类[图像分类]的任务(https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/models/image_classification/overview.md)形成对照。
当前的实现包括以下功能:
- DeepLabv1 :我们使用 atrous convolution 来显示地控制在深度卷积神经网络中计算特征响应部分的分辨率。
- DeepLabv2 :我们应用 atrous spatial pyramid pooling(ASPP) ,使用多个采样率和有效视野的过滤器,在多种尺度上稳健地分割目标对象。
- DeepLabv3 :我们使用图像级特征[5,6]来扩展ASPP模块以捕获更长距离的信息。我们还增加了批标准化[7]参数以加快训练。特别的,在训练和评估期间我们应用 atrous convolution 来提取不同输出步幅的输出特征,这在输出步幅等于16时有效的促进了批标准化训练,并在输出步幅为8时得到了更高的评估效果。
- DeepLabv3+ :我们扩展了 DeepLabv3 ,增加了一个简单但有效的解码器模块,以优化细分结果,尤其是沿着对象边界。此外,在这种编码器-解码器结构中,可以通过 atrous convolution 任意地控制所提取的编码器特征的分辨率,以折衷精度和运行时间。
Example output
该模型将以很高的精度在目标对象上创建掩膜。

Read more about segmentation
Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.
Last updated 2022-06-07 UTC.
[{
"type": "thumb-down",
"id": "missingTheInformationINeed",
"label":"没有我需要的信息"
},{
"type": "thumb-down",
"id": "tooComplicatedTooManySteps",
"label":"太复杂/步骤太多"
},{
"type": "thumb-down",
"id": "outOfDate",
"label":"内容需要更新"
},{
"type": "thumb-down",
"id": "translationIssue",
"label":"翻译问题"
},{
"type": "thumb-down",
"id": "samplesCodeIssue",
"label":"示例/代码问题"
},{
"type": "thumb-down",
"id": "otherDown",
"label":"其他"
}]
[{
"type": "thumb-up",
"id": "easyToUnderstand",
"label":"易于理解"
},{
"type": "thumb-up",
"id": "solvedMyProblem",
"label":"解决了我的问题"
},{
"type": "thumb-up",
"id": "otherUp",
"label":"其他"
}]