I colorized some black and white movies using the trained neural network model from Zhang et al.
The movies were converted to single frames via ffmpeg, colorized and then put together. As you will see this creates a lot of changes from frame to frame and could be improved by some persistence parameters for scenes. The scripting was done in python using the example from https://github.com/richzhang/colorization. The neural network is implemented in caffe.
We wanted a simple line width estimation for a recent project. It was meant to analyze the width of actin bundles. After some searching for an existing solution to the problem I only found this paper. Usually I try to use existing programs in Matlab. It is very common for physicists to reinvent the wheel.
I prepared a few images to show how the Matlab script works. I used two different line sizes for a gaussian type brush and 4 different widths for the hard edge pen tool. It should be fairly customizable to diffent approaches.
Differently oriented lines with brush size 20
I recently found out that one of my throwaway email-accounts was in the Adobe password leak. I wanted to see what was out there and downloaded the leaked data. It looks like this:
<some ID>-|-<username, mostly missing>-|-<email>-|-<password hash>-|-<password hint>|-- 103238705-|--|[email protected]|-BB4e6X+b2xLioxG6CatHBw==-|-boyfriend|-- 103238706-|--|[email protected]|-Cm8mAzxAiwzioxG6CatHBw==-|-dance|-- 103238707-|--|[email protected]|-n+TZlu41zyHioxG6CatHBw==-|-|-- 103238708-|--|[email protected]|-FAniAwP+U13ioxG6CatHBw==-|-|-- 103238709-|--|[email protected]|-kxiV+a47bSlf+E5Ulu/AzA==-|-newest|-- 103238710-|--|[email protected]|-UimSy9NunUU=-|-dog|--
Here's the relevant xkcd
I have my blog hosted on openshift using django. This is how i back up a the postgres database, copy it to my local computer, and restore into a local database. The backup uses this snippet which u can use as a cron job. I ssh'ed into my app:
rhc ssh myappname
executed the lines from this snippet: