Skip to content

frankplus/tof-stereo-fusion

Repository files navigation

ToF-Stereo Sensor Fusion

This is a project for a direct sensor fusion of two range imaging cameras:

  • A stereo vision system
  • A Time-of-Flight range imaging camera

How is it done

A convolutional neural network was designed and trained for the sensor fusion process. The model was developed in keras on Tensorflow backend

The CNN architecture

CNN architecture The CNN used for this work consists of a residual neural network with dilated convolutions. As a result of some experiments, these architectural designs have shown a positive impact on the model's performances.

Some results

In the following pictures some experimental results are shown. Column 1 and 2 represents ToF and Stereo inputs respectively. Column 3 represents the network's output. Column 4 represents the ground truth data. test 1 test 2 test 3 test 4

About

ToF-Stereo Sensor Fusion with Deep Learning

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages