引导式测验 · 中等 等级
光学字符识别测验
查看 OCR 文本转换、序列识别、预处理、Tesseract、手写限制和文档级视觉模型。
问题 1 的 8
OCR的核心工作是什么?
预览每个问题8 问题
当您参加测验时,就会出现答案和解释。
- OCR的核心工作是什么?
- 将文本图像转换为机器可读的字符代码
- 压缩图像以减少存储空间
- 自动模糊文档中的敏感文本
- 将口语音频翻译成书面字幕
- Which architecture lets OCR recognize a sequence without explicit character segmentation?
- A fixed one-template-per-character lookup
- A sequence recognizer trained with CTC or an attention decoder
- A region-proposal network only
- A color threshold rule
- 为什么“去偏斜”是 OCR 管道中常见的预处理步骤?
- 它为黑白扫描添加颜色
- 它对文档进行加密以保护隐私
- 它可以拉直倾斜或旋转的文本,使其更易于阅读
- 它将文件的分辨率提高到超出原始分辨率
- 以下哪一个是广泛使用的开源 OCR 引擎?
- 张量板
- Photoshop
- FFmpeg
- 超立方体
- 为什么手写文本通常比打印文本更难进行 OCR 识别?
- 不同作者的字母形状差异很大,并且经常连接或重叠
- 手写总是使用不寻常的墨水颜色
- 手写体总是比印刷体大
- 相机无法对焦手写页面
- What does the OCR-free Donut model do in document understanding?
- It only de-skews images and never reads text
- It processes a document image directly to produce task-specific text outputs without a separate OCR stage
- It requires Tesseract output as its input for every task
- It detects motion vectors between consecutive document frames
- What is CTC useful for in OCR?
- Encrypting scanned pages
- Correcting every spelling error
- Training sequence predictions when input-frame and output-character alignment is not given
- Finding legal signatures
- Why should OCR output be checked against the source image?
- OCR always omits spaces
- The source image is never available
- Text recognition is legally inadmissible
- Recognition can still err on degraded, unusual, or handwritten text