Skip to content

Collection of tutorials and scripts working towards and end-to-end CNN for control of microrobots.

Notifications You must be signed in to change notification settings

PisterLab/low-power-cnn

 
 

Repository files navigation

low-power-cnn

This directory is very much in development and will include a series of tutorials for undergrads supporting data collection and analysis for training an end-to-end convolutional nueral network for microrobot control. Expect description of tutorials to be added along with our new implementations. Contact: Nathan Lambert & Lydia Lee.

Research Direction

We want to equip our microrobots with control policies to explore their environment and eventually to accomplish tasks. An initial goal is to ant like cognition, by getting a small rc car or walking robot to wander the lab and avoid objects. The initial exploration is to train SqueezeDet on our own data. From there, we will need to decrease computational operations and maintain performance.

Tutorials

This section is intended for our undergrads supporting the project. The initial tutorial will walk you through the overall structure of using Tensorflow for convolutional nueral networks. Future tutorials will be added below, so turn on notifications for updates.

Also in this directory is some Pytorch Tutorials and source code. We pulled the raw code for SqueezeNet to show how custome nets are implemented. Also, we changed the Pytorch transfer learning tutorial to use SqueezeNet rather than the built in ResNet18.

For the undergrads, we will be challenging you to identify the decade of the yearbook images found here. We have prepared the data in the common structure of '~\dir\class' where the classes are decades here.

About

Collection of tutorials and scripts working towards and end-to-end CNN for control of microrobots.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 75.5%
  • Python 24.5%