18 Application example photo OCR
18-1 Problem description and pipeline
The photo OCR problem
1.Text detection
2.Character segmentation
3.Character classification (recognition)
4.*Spelling correction
Photo OCR pipeline
18-2 Sliding windows
Text detection | Pedestrian detection
Supervised learning for pedestrian detection
Sliding window detection
Text detection
1D Sliding window for character segmentation
18-3 Getting lots of data: Artificial data synthesis
Character recognition
Artificial data synthesis for photo OCR
Synthesizing data by introducing distortions
Discussion on getting more data
-
Make sure you have a low bias classifier before expending the effort(Plot learning curves). E.g. keep increasing the number of features/number of hidden units in neural network until you have a low bias classifier
-
How much work would it be to get 10x as much data as we currently have?
-
Artificial data synthesis
-
Collect/label it yourself
-
Crowd source"(E.g. Amazon Mechanical Turk)
-
18-4 Ceiling analysis What part of the pipeline to work on next
Estimating the errors due to each component(ceiling analysis
What part of the pipeline should you spend the most time trying to improve?
Another ceiling analysis example
Face recognition from images (Artificial example
19 Conclusion-Summary and Thank you
Summary: Main topics
Supervised Learning
-Linear regression, logistic regression, neural networks, SVMS
Unsupervised Learning
-K-means, PCA, Anomaly detection
Special applications/special topics
-Recommender systems, large scale machine learning
Advice on building a machine learning system
-Bias/variance, regularization; deciding what to work on next: evaluation of learning algorithms, learning curves, error analysis, ceiling analysis