This project deals with handwritten color digits by performing two tasks at a time i.e by predicting the value of the digit as well as the color of the digit. I created a data generator function that generates red, green color images using the greyscale MNIST images dataset from Keras. Here, I used Resnet style architecture by using skip connection with Add Layer in order to create a multi-tasking CNN model. The generated color images are fed to the model that is done in multi-tasking mode by implementing two CNNs for digit and colour recognition at an accuracy of 98%.
krisshvamsi / multi-task-modeling-on-hand-written-digits-using-keras Goto Github PK
View Code? Open in Web Editor NEWThis project deals with handwritten color digits by performing two tasks at a time i.e by predicting the value of the digit as well as the color of the digit. I created a data generator function that generates red, green color images using the greyscale MNIST images dataset from Keras. Here, I used Resnet style architecture by using skip connection with Add Layer in order to create a multi-tasking CNN model. The generated color images are fed to the model that is done in multi-tasking mode by implementing two CNNs for digit and colour recognition at an accuracy of 98%.