After being with Tim Weyrich’s group for almost a year and a half yesterday we had our final group lunch and a cheeky beer. It has been great working with everyone at UCL, not just in the immediate group and in that vain more shenanigans to come.

Original Image from BSDS500 Dataset

Style transfer has become a popular area of research and with public applications such as Prisma based on Neural Style Transfer [Gatys'15]. Earlier this week I wanted to answer the question does it really work for understanding the larger style and context. In contrast to [Wang'13] where they learnt an artistic stroke style, how does it compare. The British Library Flickr 1m+ dataset [BL'13] provides an interesting application of this, where there is an inherent style for transferring.

So having read the papers around this previously, I was fairly sure it would not transfer very well, but on the chance that local statistics can enforce something coherent it was worth running (and also gave me a chance to play with such networks). So by taking a few examples, these are the best results from transferring from the BL'13 to the Berkeley Segmentation Dataset (BSDS500) [BSDS500'11]

These results were achieved using the Texture Nets method [Ulyanov'16] trained on a singular example and the most visually appealing results displayed after playing with the Texture / Style weights as well as the chosen example. Code available on GitHub

What is quite interesting is if you look at this from a distance they look plausible, but it isn't till you look at one adjacent to another or zoom in you realize that these aren't actually the same style. Logical hatching patterns to describe shadows or depth are ignored, or in the cases where they are present they don't make sense.

As a mini-conclusion, style-transfer, although gaining a lot of hype has still a long way to being accurately reproduced general artistic style. Still the results are interesting and if you aren't looking for exact replication, then it is visually appealing. It must be bared in mind that the British Library dataset is challenging, where the style has been evolving for human understanding over millenniums. A problem to keep working on, possibly guided by transfer learning.

References

[Gatys'15] Leon A Gatys and Alexander S Ecker and Matthias Bethge. "A Neural Algorithm of Artistic Style". Arxiv (http://arxiv.org/abs/1508.06576). 2015.
[Wang'13] T Wang and J Collomosse and D Greig and A Hunter. "Learnable Stroke Models for Example-based Portrait Painting". Proc. British Machine Vision Conference (BMVC). 2013.
[BL'13] British Library Flickr. https://www.flickr.com/photos/britishlibrary/. 2013.
[BSDS500'11] Berkeley Segmentation Dataset. https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html#bsds500. 2011
[Ulyanov'16] Dmitry Ulyanov and Vadim Lebedev and Andrea Vedaldi and Victor Lempitsky. "Texture Networks: Feed-forward Synthesis of Textures and Stylized Images. Arxiv (http://arxiv.org/abs/1603.03417). 2016.

As I continue my crazy running activities I take another shot at the Relish Run Cheddar Gorge half marathon. A great route that takes you up over the hills surrounding Cheddar and past the trig point with views over Weston-super-Mare, my home town.

So this is a hard run and this years challenge was water, to note last years was sun. So I've tried this course into polar conditions (for the UK), it seems that water is more challenging. This was not aided by me getting lost and taking another guy with me.

Well another year another race attempt, hoping for better next year!

The San Francisco Marathon

I've been running for a few years now, doing as many runs as I can a year. Granted my training routine isn't very vigorous, but just doing the run is great fun. After attending SIGGRAPH this year, I passed through San Francisco on my drive up the coast. As any avid runner would, if there a race you can do you do. So I ran the San Francisco Second Half Marathon!

I wish I could say this went well, but this is a challenging route, but was so good to run. I hope in future I can come back and do the marathon in full, one of the unique things about this marathon it isn't just two laps of the half. Therefore by doing the marathon you get great views of the golden gate bridge and through the centre of the city.

Sam Jordan -- Flickr

When I was younger I never got into reading novels sadly. I have tried throughout my adult life to try to get into them, but sadly it is always a challenge to stick with them. Often I find the demands of my work a massive distraction result in me not wanting to sit and read more. I dabbled into Audiobooks that I recommend massively, but loosing my commute broke this. So I'm going to try listen while cycling to work!

Having spent a while now (6months) trying to finish NPCs by Drew Hayes a great fun book. Sadly the inprogress nature of this preculudes me from being able to move onto the second in the book series. Therefore my book to try next is Critical Failures by Rober Bevan. The audiobook at 8hrs should be an easy 'read' in a week, lets see if I crash!