Query by Semantic Sketch
This approach addresses the need for more intuitive and complementary query methods in multimedia retrieval, particularly for users seeking to find visual content based on semantic and spatial relationships, though it appears incremental as it builds on existing sketch-based and semantic techniques.
The paper tackles the problem of retrieving images and videos by allowing users to sketch concept maps with spatial distributions of semantic labels, merging sketch-based and semantic retrieval. The result is an efficient vector representation that enables effective retrieval in large multimedia collections, as demonstrated in the vitrivr engine.
Sketch-based query formulation is very common in image and video retrieval as these techniques often complement textual retrieval methods that are based on either manual or machine generated annotations. In this paper, we present a retrieval approach that allows to query visual media collections by sketching concept maps, thereby merging sketch-based retrieval with the search for semantic labels. Users can draw a spatial distribution of different concept labels, such as "sky", "sea" or "person" and then use these sketches to find images or video scenes that exhibit a similar distribution of these concepts. Hence, this approach does not only take the semantic concepts themselves into account, but also their semantic relations as well as their spatial context. The efficient vector representation enables efficient retrieval even in large multimedia collections. We have integrated the semantic sketch query mode into our retrieval engine vitrivr and demonstrated its effectiveness.