Machine learning
- 1. Machine Learning (OCR) Presented By- Raunaq Kataria Amit Gupta 11/29/2014 1
- 2. Overview 1. What is it ? 2. What are the Applications ? 3. What are the types of ML Algorithms ? 4. Why Machine Learning ? 5. What is our Aim ? 6. OCR Footer Text 11/29/2014 2
- 3. What is Machine Learning ? • Study of Algorithms that can ‘learn from data’ through iterations (experience) • Without being ‘explicitly programmed’ • Operation: 1. Build models on Inputs. 2. Making Predictions 3. E.g. Predicting Flight Delays Footer Text 11/29/2014 3
- 4. Applications • Spam mail (recognition) • Web click data (recommendations) • Security (Pattern recognition, face detection) • Business (Stocks, user behaviors) • Medical (Research on medical records) Footer Text 11/29/2014 4
- 5. Types Of Algorithms • There are mainly 2 types of ML Algorithms: 1. Supervised 2. Unsupervised Footer Text 11/29/2014 5
- 6. Why Machine Learning? • Beginning towards the aim of Ultimate A.I. i.e. 1. Self Aware 2. Fully Conscious 3. As Intelligent as The Human Race. Footer Text 11/29/2014 6
- 7. Our Aim • Implementing Machine Learning Algorithms to execute : OCR : Optical Character Recognition using Supervised Machine Learning Algorithm Footer Text 11/29/2014 7
- 8. What is OCR ? • OCR : 1. First step in the field of Computer Vision 2. A part of AI • Conversion of images of Handwritten or printed text into machine encoded text • Implement on MATLAB through supervised learning algorithm Footer Text 11/29/2014 8
- 9. INPUT TYPEWRITTEN TEXT IMAGE OUTPUT MACHINE ENCODED TEXT Footer Text 11/29/2014 9
- 10. Preprocessing • converted Pixel Intensities (in Matrix Forms) into two matrices Large no of training examples (say: 5000) X : rows = no. of training examples columns = pixel intensities Y : rows = no. of labels under which the training examples are classified Training Data Set Footer Text 11/29/2014 10
- 11. TRAINING DATA SET Output LEARNING ALGO (logistic regression) Hypothesis Function (h) Input Data Set < Partial Output ‘n’ iterations
- 12. OCR - PROSPECTS • Extract key info from business documents, e.g. insurance policy, passport, bank statement etc. • Automatic number plate recognition • Assistive technology for visually impaired users ( text-to-speech converter) • Extracting business card info into a contact list Footer Text 11/29/2014 12
- 13. BIBLIOGRAPHY • Lectures by - (Andrew Ng, Machine Learning- Stanford University ) • http://en.wikipedia.org/wiki/Machine_learning Footer Text 11/29/2014 13
- 14. Thank You Footer Text 11/29/2014 14