Research in computer vision for recognizing 3D objects in photographs began in the 1960s, but significant advancements did not occur until the early 2000s. Recently, effective techniques have emerged for recognizing object categories despite variations in observation conditions. The last five years have seen remarkable progress, largely due to the integration of new data representations, like invariant semi-local features, from the computer vision community with robust models of data distribution and classification from the statistical machine-learning field. This volume contains selected papers from two workshops held in Taormina in 2003 and 2004, which aimed to foster an international object recognition community with shared datasets and evaluation methods. The workshops focused on mapping the state of the art, identifying key challenges and opportunities for collaborative research, and addressing industrial and societal needs in object recognition. The 30 revised papers are organized into sections covering specific object recognition, object category recognition, recognition involving geometric relations, and joint recognition and segmentation.
