Advanced Multimedia Processing (AMP) Lab, Cornell University

Pictorial Structures for Object Recognition and
Part Labeling in Drawings



Amir Sadovnik, Tsuhan Chen


Although the sketch recognition and computer vision communities attempt to solve similar problems in different domains, the sketch recognition community has not utilized many of the advancements made in computer vision algorithms. In this paper we propose using a pictorial structure model for object detection, and modify it to better perform in a drawing setting as opposed to photographs. By using this model we are able to detect a learned object in a general drawing, and correctly label its parts. We show our results on 4 categories.

More specifically, the input to the algorithm is a general drawing made up of separate strokes (Fig. 1(a)). In each of the drawings, one or more objects of different catagories are drawn, in addition to strokes which do not belong to any known object. An object detector for each category is then run on the drawing. The output of each detector is a new image, in which each stroke is labeled as part of the object, or as background (Fig. 1(b)). Each detected object also receives a confidence score. Objects with low confidence are removed from the output.


Amir Sadovnik, Tsuhan Chen. Pictorial Structures for Object Recognition and Part Labeling in Drawings. International Conference on Image Processing (ICIP 2011).


The dataset used will be made available soon on this page to download for further research.

Fig 2. Examples of inputs and results obtained by our different detectors. Arrow colors signify which detector has been used. (Green-Body, Yellow-Face, Red-Flower, Blue-House). Click image to enlarge.