Teaser: Sketch-based image retrieval


We introduce a benchmark for evaluating the performance of large scale sketch-based image retrieval systems. The necessary data is acquired in a controlled user study where subjects rate how well given sketch/image pairs match. We suggest how to use the data for evaluating the performance of sketch-based image retrieval systems. The benchmark data as well as the large image database are made publicly available for further studies of this type. Furthermore, we develop new descriptors based on the bag-of-features approach and use the benchmark to demonstrate that they significantly outperform other descriptors in the literature.

Benchmarking your sketch-based retrieval system

  1. Download the benchmark dataset. This dataset contains 31 benchmark sketches as well as 40 corresponding images for each sketch. It also includes the complete results from the user study described in the paper as well as Matlab code to evaluate your system's results against the benchmark defined in the paper.
  2. Download the distractor image dataset.
  3. Copy the benchmark images contained in the benchmark dataset to the folder containing the distractor images.
  4. Run your sketch-based retrieval system with each of the 31 benchmark sketches as the query. For each query, store a list that contains the ranking of the corresponding 40 benchmark images. The order of the results in these lists is essential and is defined in the Readme file that comes with the dataset.
  5. Use the resulting 31 lists to evaluate your system with the given Matlab code.



author = {Eitz, Mathias and Hildebrand, Kristian and Boubekeur, Tamy and Alexa, Marc},
title = {Sketch-Based Image Retrieval: Benchmark and Bag-of-Features Descriptors},
journal = {IEEE Transactions on Visualization and Computer Graphics},
year = {2011},
volume = {17},
number = {11},
pages = {1624--1636}