Deep time (using Tensorflow to read clocks)

This is a simple project with the aim of learning to tell time from an image of a clock like this one:

The code is available here: https://github.com/felixduvallet/deep-time-reading

In a nutshell, the approach looks something like this:

• Generate a synthetic clock face with different times shown (input data).
• Use tensorflow batch queues to provide a parallelizable input pipeline.
• Formulate this problem as a multi-task classification problem, where one task is to predict the hour and one task is to predict the minute.
• Use (essentially) CIFAR10’s model architecture.
• Optimize cross entropy loss
• Also measure prediction precision and time error (how far off our time telling is from the truth).

glances, an awesome system monitor.

I found a really nice system monitor called Glances:

Installation is super simple:

$ sudo pip install glances
$ glances

You can also run this in web mode, which is useful for visualizing the system load on a remote server somewhere:

$ glances -w

Then, just point your browser to http://@server:61208

NOTE that you need to have bottle installed: $ sudo pip install bottle.

Unittesting ROS in python

Setting up unit tests for ROS python modules is a little tricky, but once it’s in place you can easily automate testing of ros nodes.

Here is a very brief reference.