-
-
Notifications
You must be signed in to change notification settings - Fork 1.5k
Open
Labels
Accepted 🌱 受理Maintainer may be busy resolving or fixing the labeled issuesMaintainer may be busy resolving or fixing the labeled issuesBug 🐞 故障Confirmed bugs or reports that are very likely to be bugsConfirmed bugs or reports that are very likely to be bugsFeedback Awaited 🕒 等待反馈Waiting for the feedback from issue creater and issue will be closed if duration is over one monthWaiting for the feedback from issue creater and issue will be closed if duration is over one monthProposal 💡 方案Issue creater's own proposal which may be really helpfulIssue creater's own proposal which may be really helpful
Description
- 环境信息
- 设备型号: (红米k80pro)
- Android 版本: (例如: os版本:3.0.3.0)
- Auto.js 6 版本: 6.6.4
- 问题描述
根据官方文档,ocr.mlkit.detect() 和 ocr.paddle.detect() 应该分别调用 MLKit 和 PaddleOCR 两个不同的识别引擎,并产生不同的识别结果。
然而,在我的环境中,无论调用 ocr.mlkit.detect(img) 还是 ocr.paddle.detect(img),两者返回的识别结果(包括所有文本块的 label, confidence 和
bounds)都完全一致,精确到每一个细节。这表明底层实际上可能只调用了同一个 OCR 引擎,或者其中一个引擎静默失败并回退到了另一个。
-
预期行为
调用 ocr.mlkit.detect(img) 和 ocr.paddle.detect(img) 时,应该分别激活两个独立的 OCR 引擎,并针对同一张图片返回两组具有明显差异(尤其是在置信度 confidence 上)的识别结果。 -
实际行为
两个函数返回了完全相同的 OcrResult[] 数组,使我无法对两个引擎进行独立的测试和使用。
Metadata
Metadata
Assignees
Labels
Accepted 🌱 受理Maintainer may be busy resolving or fixing the labeled issuesMaintainer may be busy resolving or fixing the labeled issuesBug 🐞 故障Confirmed bugs or reports that are very likely to be bugsConfirmed bugs or reports that are very likely to be bugsFeedback Awaited 🕒 等待反馈Waiting for the feedback from issue creater and issue will be closed if duration is over one monthWaiting for the feedback from issue creater and issue will be closed if duration is over one monthProposal 💡 方案Issue creater's own proposal which may be really helpfulIssue creater's own proposal which may be really helpful