Tensorflow apk for THETA V - Speech, object detection, stylized filters

tensorflow

#1

Speech detection, object detection, stylized filters, deep learning classification all working inside the THETA V. Download pre-compiled Tensorflow apk for developers or power users with developer mode enabled.

image

image

image

image

image
image

image


#2

Wow, this is cool! I mean, SERIOUSLY cool! I grabbed the apk that @codetricity built and installed it in my THETA V. Very smooth. For once, I remembered to set the permissions properly (only need to do that once, the first time it’s installed) and was able to run the plug-in from Vysor very easily. Here’s screenshots from two of the four functions.

Object Detection

Apparently Tensorflow has 95% certainty that I’m a person. I feel good about this.

2018-11-07-1

Stylized Filters

I could probably spend some time playing with this! There appear to be at least 26 filters included in the plug-in by default.

2018-11-07


#3

Fantastic. Thanks for posting.

I think it would be useful to show more examples of use. Consider making a video of your tests with TensorFlow.

The background research for the artistic styles is found in this paper.

This image shows you the basic idea for the artistic style training.


Feel free to post more tests.


#4

I added more information on how to download the TF Stylize images from the RICOH THETA V.


#5

One of the problems with equirectangular images is that a standard object dataset like COCO (Common Objects in Context) doesn’t play well with objects on the edge.
This paper talks about using COCO with YOLO.

Their solution is to use four sub-windows.

As this is a research paper, it seems unlikely we’ll be able to solve the problem with the THETA V to get full 360 object detection. We may be able to a reasonable 210 degrees of detection without too much work. I’m curious to learn how other people have progressed.


#6

I was thinking about this overnight and I remember reading through the Automatic Face Blur code and seeing that the equirectangular image was divided into four parts.

At the time, I was wondering why the developer divided the image into four parts. Now, I’m wondering if it could be related.