For some time, I have been into entrepreneurship. More recently, I have become a better programmer while doing my PhD at Cambridge. This has led to http://www.rasterfarian.co being launched recently.
Rasterfarian is a photo enlargement web-based application.
It started from two directions. One was trying to apply image processing techniques I’ve learned in my PhD to something more widely applicable. The second was the need to expand small photos without losing quality.
My department has a small logo, which as far as I know only exists on its website. I had wanted to put it on a poster presentation, but blowing it up ~20X made it look terrible – really pixelated. I wondered if there was a way to enlarge it without losing so much quality.
It turns out, there is. The general method I use is similar to bicubic interpolation (http://en.wikipedia.org/wiki/Image_scaling#Algorithms) – but with a few refinements and improvements. I use Python’s Scikit-Image Library (http://scikit-image.org/) to power much of the algorithm, and host on Red Hat Openshift (www.openshift.com). One day I’ll write a post about the implementation of everything.
Traffic has been interesting to the site. The first wave was from friends on Facebook and Twitter. This traffic came quite quickly, and crashed my server a few times so I upgraded. I made some improvements to file handling, scaling, and the UI from feedback from this round.
The second stage of traffic was from Hacker News. This again was a quite strong pulse of traffic. My servers stayed up for this, and I made further improvements to the UI from feedback from HN users. In particular, I improved the site’s mobile appearance and started using Bootstrap. For much of my early development, I didn’t test as much as I should have, especially with different devices and browsers.
More recently, I’ve been getting a lot of my traffic from Google Adwords. This has been quite interesting (and thank you Google for ad credits for start-ups – I wouldn’t have stuck with you this long without them!). Many of my users are from India, the Philippines, and Romania. Its exciting that people all over the world are using my app, but I hadn’t expected this. My page is in English, and I don’t know if its translated or easily usable for speakers of other languages. I had put ads on my site to see if they would be profitable – but the are targeted at a USA and UK audience.
Google Adwords seem to direct traffic to my site, and when linked to analytics they optimise the traffic well to users that convert. The optimisation takes a few days, and changing parameters in Adwords seems to break the optimisation each time. I’ve learned to avoid reacting or experimenting too quickly – it takes a while to get proper results. I’ve used Bing Adwords as well and don’t seem to get as good of results, but this could be that I’m tracking with Google Analytics 🙂
The ad on my site is through an affiliate marketing program. I have a decent click-through rate, but no conversions so far. I am waiting to collect enough data, but I’m not sure this model works for my site. I had hoped to create a small amount of income to have some greater freedom and independence, not a major venture-funded business (on this idea!). So far, this model doesn’t seem to be working. Further, if many of my visits are from Google Adwords, its like I’m paying for people coming in, hoping they will just leave. Should I target visitors to my ad or to use my site? These may be conflicting goals.
I plan to write a longer post on “Problems you think you have” from some of my experiences with http://www.rasterfarian.co
When making the site, I’d worried about things that were lower priority or in the distance. How do I scale to hundreds of users on the site at a time (I typically have only hundreds per day, several weeks after launch)? How do I remove all the data my users will be generating? Instead, I should have tested more immediate things like UI/ UX on mobile devices and IE.
There continues to be a long list of updates I’m hoping to add to Rasterfarian. I also have several other ideas for small apps I plan to launch, all in the computer-aided design space. Rasterfarian and these other apps will be under the Makeraas (Maker As A Service) brand. Makeraas is a company I’ve started to democratise computer aided design (CAD) software. CAD for engineering and artists has become quite specialised and difficult to use – but many final applications like 3D printing and digital photo sharing has never been easier. My goal is to lower the learning curve in using CAD to make these new technologies more accessible. Rasterfarian, for example, enlarges photos in a way that is relatively easy in Photoshop or GIMP, but most people don’t have the time to learn the software.
Stay tuned for updates!