It doesn’t save, doesn’t speak, doesn’t fall. It floats in the void of code.

Another generative painting based on an awesome NASA photo. Unretouched output from my software, except for sharpening and some text, added in PhotoShop.

This one is made up of of two layers:

1) A painted under layer

2) A drawing layer to add detail

By adding a drawing layer to the painting, I'm trying to make the image less photo/filter-like.

Drawing, is a much harder problem to solve algorithmically than painting. It's very easy to 'cheat' when painting -- there are a lot of color and optical effects that you can put in between the subject and the canvas.

With drawing -- in the traditional sense -- it's more about understanding the form and putting that down as simply as possible, than just copying what you see. Shapes and edges that get lost in a painting instead need to be disambiguated. This is very difficult for the computer to do because it doesn't have an understanding of what it's looking at.

Check out the original size image, if you can, the surface feels good on this one. There are some nice details in the brushwork, I think.

Here's the original photo (GPN-2000-001027): dayton.hq.nasa.gov/ABSTRACTS/GPN-2000-001027.html

Follow me on twitter: Tinrocket

Providing the work can be referenced from two points (a datum and a reference) and in this case, using a small webcam, it is two small holes along one edge then it does not have to be exactly or squarely placed on the work table in order to be accurately machined.

A short video. www.graytel.talktalk.net/rotate1.wmv

For any that are interested there are more details of how I am using the camera here; hobbycncart.com/publ/cikkek/mach3_temaju_cikkek/tool_posi...

Nils Peek van DVC Machinevision en Jan Zwaan van Espera.

Inspectiesysteem van DVC Machinevision gekoppeld aan een Nova weeg- en etiketteersysteem van Espera.

On the left, a webcam observing a game of Connect 4. On the right, a window showing that the computer has recognised the game board and also knows whether each cell has a red piece, yellow piece or is empty.

Just above all that is a robot arm that can pick up pieces from its hopper and play a game against you.

Here are some fun things to try to see just how smart it is.

- move the game board a little

- move the camera

- put a red token in the yellow token hopper; will the robot avoid dropping the token when it sees it's the wrong colour, or will it drop it anyway? How confused will it get if it sees itself place an opponent piece?

An experiment in found machine-vision footage: vimeo.com/36239715

HDR of a part conveyor feeding a Fanuc 2000i robot. Above the conveyor is a vision sytem with infared led lighting detecting correct part orientation before the parts get to the robot.