Lab Assignment 8 MNIST and Google CardBoard App - rashmitripathi/Big_Data_Analytics_And_Apps GitHub Wiki

1. Write a TensorFlow program for the following Task.

a. Implement SoftMax Classification for Image Dataset that is not covered in class. Report accuracy,

b. Visualizations(Tensor Board): training and testing both.

MNIST is a simple computer vision dataset. It consists of images of handwritten digits.It also includes labels for each image, telling us which digit it is.As part of this project, I have selected the dataset as shown in the figure below. It contains of digits 0-9 and then MNIST regression is done on the same.

The following steps are followed:

  • Create a softmax regression function that is a model for recognizing MNIST digits, based on looking at every pixel in the image
  • Use Tensorflow to train the model to recognize digits by having it "look" at thousands of examples (and run our first Tensorflow session to do so)
  • Check the model's accuracy with our test data

Then I have calculated the training accuracy and testing accuracy.

Testing accuracy is 97.53 %

Visualizations for Testing and Training data:

For Accuracy

For Cost

Cross_Entropy

Tensorflow Graph

Histograms

2. Develop a Cardboard App that is relevant to your own project

360 Video Viewer with an additional feature such as

360 Video viewer:

a. Spatial audio:

i. https://developers.google.com/vr/android/spatial-audio

ii. https://developers.google.com/vr/ios/ndk/reference/group/audio

b. Head tracking.

i. https://developers.google.com/vr/ios/ndk/reference/group/headtracking

c. User input event handling.

i. https://developers.google.com/vr/android/samples/treasure-hunt#handling_inputs

d. Game:

i. https://developers.google.com/vr/android/samples/treasure-hunt

I have created the demo of treasure hunt as shown in screen shots below:

References: