pre-processing images using octave for computer vision project -my first github project tomi
The images are hand-drawn alphabetic letters from A to J.
I developed functions for pre-processing images before feeding to a neural network to train a model using TensorFlow. The images were hand-drawn by me, which is not an effective way of getting training examples. Therefore for creating more training examples from existing images use the datasetmaker.m The model made predictions on the cross-validation set with 85% accuracy. The cross-validation set was kept away from the model therefore it had never been seen before
Code for training a model found in TensorFlow training of model.ipynb
To create new images use function datasetmaker.m it requires as input (number of images, pixel by pixel , 1==left or 0==right)
To create matrix of images with each row an example and each column a feature use function imagematrix.m it requires as input (number of images, pixel by pixel)
For feature scaling and normalization use function trainsetnormalize.m for the training set and testsetnormalize.m for the test set
- Ensure images you are working with are in current directory
- datasetmaker2.m and imageprocessor.m are helper functions ensure they are in the current directory
- Images need to be named numerically to utilize functions e.g The first image will be named 1.png