Optical Character Recognition Quiz

Answers and explanations appear as you take the quiz.

1. What is the core job of OCR?
   • To convert images of text into machine-readable character codes
   • To compress images so they take less storage space
   • To blur sensitive text in documents automatically
   • To translate spoken audio into written subtitles
2. 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
3. Why is 'de-skewing' a common preprocessing step in OCR pipelines?
   • It adds color to black-and-white scans
   • It encrypts the document for privacy
   • It straightens tilted or rotated text so it is easier to read
   • It increases the file's resolution beyond the original
4. Which of these is a widely used open-source OCR engine?
   • TensorBoard
   • Photoshop
   • FFmpeg
   • Tesseract
5. Why is handwritten text generally harder for OCR than printed text?
   • Letter shapes vary widely between writers and often connect or overlap
   • Handwriting always uses unusual ink colors
   • Handwriting is always larger than printed text
   • Cameras cannot focus on handwritten pages
6. 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
7. 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
8. 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