A GPU Accelerated Evolutionary Computer Vision System
Eberhard Karls · 2009
We have used the graphics processing unit (GPU) of the graphics card to create an evolutionary image processing system which is able to learn how to detect a user-specified object in an image. The system receives an image sequence as input. The user only has to tell the system where this object is located. This is done by using the mouse pointer. The user simply moves the mouse over the desired object and then presses the mouse button as long as the object is located under the mouse pointer. The user follows this object over several frames while keeping the mouse button pressed. As this is being done, the system evolves a population of image processing algorithms by exploiting the power of the GPU at interactive rates. Our system is the first GPU accelerated evolutionary image processing system (Figure 1) which allows the automatic creation of object detection algorithms [2]. This is the first step towards building fully adaptive evolutionary vision systems [1]. Consumer graphics cards are specifically optimized to render images at high speeds. A three-dimensional scene consists of numerous triangles which are fed to the graphics card. In order to obtain photo-realistic images, small programs can be sent to the graphics card to specify computations which should be carried out per vertex (vertex shaders) or per pixel (pixel shaders). The OpenGL shading language (OpenGLSL) has been developed as a standard to program vertex and pixel shaders. This shading language as well as the computations which are carried out on the graphics card are highly optimized for rendering three-dimensional scenes consisting of thousands of triangles. We use this programming paradigm to perform image processing on the graphics card efficiently. To fully exploit the power of the GPU, we use exactly the same paradigm which is used when rendering images. Only a single polygon is rendered. This polygon represents the output image of the image processing algorithm. The image processing algorithm (generated by simulated evolution) is fed to the pixel shader. This pixel shader is then used to compute the correct output color for each pixel. The original input image is supplied to the pixel shader as a texture.