hand-drawings are simpler, less informative
class variability: variations among instances within an object class
Chamfer matching methods can detect shapes in cluttered images, but they need a large number of templates to handle shape variations, e.g. 1000, and are prone to produce rather high false-positive rates
a powerful point-matching method based on Integer Quadratic Programming: computational complexity
Besides,  uses real images as models, so it is unclear how it would perform when given simpler, less informative hand-drawings.
 based on edge patches
Contour Segment Network
- dealing with highly cluttered images,
- allowing intra-class shape variations and large scale changes,
- working from a single example,
- being robust to broken edges, and
- being computationally efficient
brittleness of edge detection – contour is often broken into several edgel-chains
segment the contour chains of the model, giving a set of contour segment chains along the outlines of the object
functionality of pure shape matchers – takes a clean shape as input, support matching to cluttered test images
Simple decomposes the hand-drawing into PAS, then uses these PAS for the Hough voting stage, and the hand-drawing itself for the shape matching stage.
representative shape context
- shape context of which point should represent the image?
- pixel density based sampling – promote point with higher or lower density?
sampled shape context of the shape may contain redundant information
Mori et al.  tested the representative shape context method on the Snodgrass and Vanderwart line drawings.
- Queries were distorted versions the original
Embed objects into some clutter:
find the outline of the object, construct a binary mask for it, and using logical operations (AND à OR) to copy the clutter around the object.
- Finding the outline of objects is done using a method similar to flood-fill.
- Pseudocode for original Representative Shape Context
% Compute shape contexts for known shapes
SCquery = shape contexts for r random pointsforeach known shape Si
for j = 1 : r
dist(Squery; Si)+ = minu(2(SCj query; SCui ))
% Sort dist and truncate to return a
A query of a hand-drawn shape is successful if the corresponding known shape is included in the set of retrieved candidate shapes.
‘Shape Context and Chamfer Matching in Cluttered Scenes’ – only a single template shape
paper From Images to Shape Models for Object Detection [pdf]
Hierarchical Matching of Deformable Shapes[pdf]
 Mori, G., Belongie, S., & Malik, J., (2001) Shape Contexts Enable Efficient Retrieval of Similar Shapes, CVPR.[pdf]
 Recognizing hand-drawn images using shape context [pdf]