AI入门之——Andrew Ng “Machine Learning”课程学习笔记第十一周

11、Application Example: Photo OCR

11.1、Photo OCR

11.1.1、problem description and pipeline

photo OCR pipeline

  1. Text detection
  2. Character segmentation
  3. Character classification

11.1.2、sliding window

对整个图片进行分别窗口化检测

11.1.3、Getting lots of data: Artificial data synthesis

Synthesizing data by introducing distortions

  1. Distortion introduced should be representation of the type of noise/distortions in the test set.
  2. Usually does not help to add purely random/meaningless noise to your data.

11.1.4、What part of the pipeline to work on next

进行分块处理的目的是,我们可以很容易的分析出那一步骤是系统性能瓶颈,需要在那一步骤上投入更多的精力。

11_1

11_2