I am not sure if there is a better forum to ask this question, so I will try here.

Consider an image, say in .jpeg format, representing a directed graph, with nodes. Is there an algorithm, be it a machine learning/image processing/computer vision algorithm, that can read and process this image, and generate a potential adjacency matrix representation of this graph? What if the graph isn't planar, i.e. edges can intersect.

The graph itself could be unshaded circles representing nodes, with an arrow representing an edge from one node to another.

Obviously, as it is a supervised learning problem, some training data needs to be provided, but if this can be done, how would I go about doing it? If someone can point me to research in this problem, I'd appreciate it, as my search seems to have been to no avail.


1 Answer 1



Christopher Auer, Christian Bachmaier, Franz-Josef Brandenburg, Andreas Gleißner, Josef Reislhuber: Optical Graph Recognition. Graph Drawing 2012, pp. 529-540 http://dx.doi.org/10.1007/978-3-642-36763-2_47


Not the answer you're looking for? Browse other questions tagged or ask your own question.