Handwritten Digits Recognition by Using CNN Alex-Net Pre-trained for Large-scale Object Image Dataset
Yoshihiro Shima, Yumi Nakashima, Michio Yasuda · 2018
Neural networks are powerful technology for classification of character patterns and object images. A new method for handwritten digits recognition is proposed by combining pre-trained Convolutional Neural Networks (CNN) and Support Vector Machines(SVM). Pre-trained CNN, Alex-Net can be used as pattern feature extractor. Alex-Net is pre-trained for large-scale object image dataset. An SVM is used as trainable classifier. The training 60k samples on MNIST database are trained by the SVM. The feature vectors of character patterns are passed to the SVM from Alex-Net. Without augmentation, experimental results of test error rate 1.03% on the test 10k MNIST database shows that proposed method is effective in handwritten digits recognition.