Data Augmentation using Tensorflow

Basically for training any Image Based Machine learning or deep learning models more number of Images Required to get better Accuracy in classification or mAp in object detection.

In Real world Scenarios Collecting Images of Some classes is a difficult task because we can’t find those images in anywhere in web. It may leads to overfitting or underfitting.

If the model is not able to classify on training images is known as underfitting which leads to more training error. One of reasons for underfitting is having less training data set size or less number of layers

If the model is not able to classify on testing images or not generalized then the model is said to be overfitting

In those Cases Data Augmentation can help a lot to increase our training or testing dataset size. It will perform various operations on single image like rotation, zooming, scaling etc which is used to create a new image from the existing

Data Augmentation Using Tensorflow:

Using Tensorflow Data augmentation is easy we just have to input an image and should set the various options in Image data generator function inside keras.preprocessing.image

A final year under-grad student and data science enthusiastic