Style Transfer Docker images

I would like to share with you a small setup I did for artistic style transfer. I won’t go too much into technical details, rather I will describe my motivations and what problems I tried to solve.

style_transfer_example

Style transfer example taken from fast-style-transfer.

Style transfer can easily produce quite interesting results unlike filters we already know from e.g. Adobe Photoshop. While style transfer might be cumbersome to set up and control, it is definitely an interesting image processing technique to experiment with and which can produce novel looks.

Continue reading “Style Transfer Docker images”

Advertisements
Style Transfer Docker images

Photogrammetry | Photos and colors

In the next couple of posts I will be describing photogrammetry workflow which I am creating for a student production. I am planning to describe here the whole process, from processing pictures to exporting game-ready assets. Game-ready assets because our production requires realtime renderings, but the process can be easily altered to produce VFX-ready assets (but this will probably require more manual work e.g. retopo, UVs to meet stricter quality requirements).

I have a tight schedule for this project, so I will try to stay simple, efficient, but try to be as correct as possible.

In this post I will describe our photos processing setup and colors workflow. This will result in photos with mask in alpha channel and in linear color space, ACES in our case. After that the photos should be ready for photogrammetry software, but that is for another post 🙂

47109051_2177820735793256_661838156617744384_n.jpg

Continue reading “Photogrammetry | Photos and colors”

Photogrammetry | Photos and colors

Nuke tip | Batch conversion of raw images with dcraw

I recently started to help out with photogrammetry pipeline for a student project. I would like to do artist-friendly yet accurate and proper workflow. It should be as automatic as possible. But that I will describe more in depth in other blog posts.

One of the first steps of our workflow is to process raw images (Canon’s .CR2 in our case) into something Nuke can handle well. In this quick tip I will show you a simple tool I adapted for this purpose and how you can use it directly from Nuke.

screen_nuke

Continue reading “Nuke tip | Batch conversion of raw images with dcraw”

Nuke tip | Batch conversion of raw images with dcraw

Nuke tip | Setting up NumPy on Windows

Sometimes it is useful to have NumPy module available in Nuke’s Python. It contains powerful tools for image processing and is required by some gizmos, e.g. mmColorTarget.

Setting up NumPy on Linux is usually just a matter of running “$ pip install numpy”, but on Windows it can get tricky.

In this quick tip I will show you how to easily take advantage of Houdini’s NumPy which was compiled with the same compiler as Python in Nuke.

Continue reading “Nuke tip | Setting up NumPy on Windows”

Nuke tip | Setting up NumPy on Windows

Precision in Houdini

In my recent project in Houdini I was often hitting precision limits of 32-bit floating point numbers. This led me to figuring out what number precisions are used in Houdini and how can they be accessed.

In this post I will take a look at various contexts of Houdini and will mention floats, but the situation for integers should be the same.

Currently it looks that single precision (32-bit) floating point numbers are used in most operations and transferring data between nodes often happens in this format too. However this situation seems to be likely changed in the future releases of Houdini.

Continue reading “Precision in Houdini”

Precision in Houdini