Hacktoberfest 2026：维护者为十月标记出来的 issue，仍然开放、适合新手。 浏览 Hacktoberfest issue

Image Comparison for Dynamic and Multilingual screen

未关闭
  • 哪里不对？
#2,426 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 5 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
25/100
Issue 类型
功能
描述清晰度
需要澄清
活跃度
冷清
技术栈
java
领域
computer-vision, testing-qa

调研方向

首先审阅 SimilarityMatchingOptions 以及链接的 Appium Images 插件文档中关于相似度计算和出现位置查找的内容。比较所请求的文本排除、阈值、坐标或元素遮罩以及基于 AI 的替代方案，然后确定一个可实现的范围和测试，以证明预期的多语言和动态文本行为。

由索引模型根据 Issue 内容生成。

描述

Hi,
I am using Appium Images plugin to perform visual testing, I have found Similarity Calculation and Occurences Lookup to be particularly useful for my use case where I am comparing if a screen matches the baseline, and if a certain element screenshot is present as a subset of screen.
For multilingual apps, I have found one limitation is that it picks up the translation differences which results in reduced value of ComparisonResult. For example,
English Screen

Image

German Screen

Image

Similarity Result

Image

In case, where the baseline screenshot is the same language (English) then the similiarity score comes out to be 0.999995768070221 while when the baseline screenshot is in a different language (German) then the similarity is significantly low 0.8476240038871765 due to the different text on image.

To counter this difference, one approach would be to run tests against different locales in which baseline images stored for comparison correspond to the lanague of app under test. While this approach is helpful, it adds an overhead to run tests on multiple lanaguages and does not scale if no of supported languages is large.

This is also relevant for cases where the text is variable for example, name/address/price of an on-screen element can differ. While mocking the result to have always same values for image comparison can be done, this is not always the possibility and easier to achieve.

For such image comparison, I would like to have a feature that basically ignores any text on the screen when performing image similiarity. Would it be possible to have such an option in SimilarityMatchingOptions where this flag can be passed to either consider or ignore on-screen text together with a threshold value which decided if and how much on-screen text will be taken into account during image similarity check? Another idea is to pass element ID or screen coordinates which should be excluded from image comparison, maybe to mask/blackout them, this way we can manually pass elements which have dynamic text to be excluded for similiarity matching?

Or perhaps if this is not a possibility, would it be possible to achieve this visual testing without relying on external providers and only using Appium MCP with some sort of image comparison tooling either using Images plugin or via LLM native image models. There is a feature request I have also put up for https://github.com/appium/appium-mcp/issues/418. There is a tool available for finding elements using AI but not sure if it can be extnded to also use AI for image comparison.

If there is a rather more suitable approach that already exists, I would be also interested in trying it out. thanks

主要语言
Java
星标
1.3k
派生
755
平均合并
5 天 14 小时
30 天内合并 PR
8

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 有 Pull Request 模板
  • 没有贡献指南

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

appium/java-client 的其他 Issue

查看 appium/java-client 的全部 Issue

相似的 Issue

更多 Java Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。