Herwig Lejsek

NV-Tree: An Efficient Disk-Based Index for Approximate Search in Very Large High-Dimensional Collections

Over the last two decades, much research effort has been spent on nearest neighbor search in high-dimensional data sets. Most of the approaches published thus far have, however, only been tested on rather small collections. When large collections have been considered, high-performance environments have been used, in particular systems with a large main memory. Accessing data on disk has largely been avoided because disk operations are considered to be too slow. It has been shown, however, that using large amounts of memory is generally not an economic choice. Therefore, we propose the NV-tree, which is a very efficient disk-based data structure that can give good approximate answers to nearest neighbor queries with a single disk operation, even for very large collections of high-dimensional data. Using a single NV-tree, the returned results have high recall but contain a number of false positives. By combining two or three NV-trees, most of those false positives can be avoided while retaining the high recall. Finally, we compare the NV-tree to locality sensitive hashing, a popular method for ¿-distance search. We show that they return results of similar quality, but the NV-tree uses many fewer disk reads.

Scalability of local image descriptors: a comparative study

Computer vision researchers have recently proposed several local descriptor schemes. Due to lack of database support, however, these descriptors have only been evaluated using small image collections. Recently, we have developed the PvS-framework, which allows efficient querying of large local descriptor collections. In this paper, we use the PvSframework to study the scalability of local image descriptors. We propose a new local descriptor scheme and compare it to three other well known schemes. Using a collection of almost thirty thousand images, we show that the new scheme gives the best results in almost all cases. We then give two stop rules to reduce query processing time and show that in many cases only a few query descriptors must be processed to find matching images. Finally, we test our descriptors on a collection of over three hundred thousand images, resulting in over 200 million local descriptors, and show that even at such a large scale the results are still of high quality, with no change in query processing time.

NV-Tree: nearest neighbors at the billion scale

This paper presents the NV-Tree (Nearest Vector Tree). It addresses the specific, yet important, problem of efficiently and effectively finding the approximate k-nearest neighbors within a collection of a few billion high-dimensional data points. The NV-Tree is a very compact index, as only six bytes are kept in the index for each high-dimensional descriptor. It thus scales extremely well when indexing large collections of high-dimensional descriptors. The NV-Tree efficiently produces results of good quality, even at such a large scale that the indices cannot be kept entirely in main memory any more. We demonstrate this with extensive experiments using a collection of 2.5 billion SIFT (Scale Invariant Feature Transform) descriptors.

Videntifier: identifying pirated videos in real-time

With the proliferation of high-speed internet access and the availability of cheap secondary storage, movie piracy has become a major problem. This demonstration paper describes the Eff2 Videntifier, a content-based system for large-scale automatic copyright enforcement of videos. The paper briefly describes the database and image processing techniques underlying the system. It also describes our proposed demonstration, which realistically simulate scenarios of copyright violations of movies.

Videntifier™ forensic: robust and efficient detection of illegal multimedia

A large portion of the video material available on the Internet is distributed illegally. In this demonstration we present Videntifier Forensic, a new law enforcement solution for automatically identifying videos and images. Videntifier Forensic is very robust and efficient, even at a very large scale. We encourage ACM Multimedia participants to bring original videos and modified (yet visually acceptable) copies to challenge the capabilities of the system.

Dynamic behavior of balanced NV-trees

In recent years, some approximate high-dimensional indexing techniques have shown promising results by trading off quality guarantees for improved query performance. While the query performance and quality of these methods has been well studied, however, the performance of index maintenance has not yet been reported in any detail. Here, we focus on the dynamic behavior of the balanced NV-tree, which is a disk-based approximate index for very large collections. We report on an initial study of the effects of several implementation choices for the balanced NV-tree, and show that with appropriate implementation, significant performance improvements are possible. Overall, the proposed techniques not only reduce maintenance cost, but can also improve search performance significantly with minimal loss of search quality.

GPU Acceleration of Eff2 Descriptors using CUDA

Video analysis using local descriptors requires a high-throughput descriptor creation process. This speed can be obtained from modern GPUs. In this paper, we adapt the computation of the Eff2 descriptors, a SIFT variant, to the GPU. We compare our GPU-Eff descriptors to SiftGPU and show that while both variants yield similar results, the GPU-Eff descriptors require significantly less processing time.

Videntifier™ forensic: a new law enforcement service for automatic identification of illegal video material

Tracking down producers and distributors of offensive video material, in particular child pornography, has become an ever growing focus of the worlds law enforcement agencies. We describe Videntifier Forensic, a new service which radically improves the forensic video identification process, by providing law enforcement agencies with a robust, fast and easy-to-use video identification system. Using this service, a single mouse-click is sufficient to automatically scan an entire storage device and classify all videos. We give an overview of the service and the underlying technology components. We then describe an acceptance test, performed by the Icelandic police forces, which demonstrates the robustness of the service.

A case-study of scoring schemes for the PvS-index

Recently we have proposed a new indexing method for high-dimensional data, the PvS-index. It provides fast query processing in constant time and is well suited for doing similarity search in Image Retrieval Systems using local descriptors. It is based on projecting data points onto random lines and uses this information to segment them into appropriately sized buckets, which can be read in just one I/O operation. After this preprocessing step the search queries just three buckets per query descriptor and uses a recent rank aggregation method, OMEDRANK, in order to provide good approximate results for the nearest neighbour problem.We have recently shown that PvS-indexing works well for large collections of real image data. In that work, however, we used a simple scoring scheme and collected few nearest neighbours for each query descriptor. In this study we examine how much the actual number of nearest neighbours, gathered for each local descriptor, influences the final query result, when searching a PvS-index. Based on the results we propose two new alternative scoring schemes, which improve the retrieval quality and stabilise the results, making the search less affected by the actual number of nearest neighbours accumulated.

Blazingly fast image copyright enforcement

Many photo agencies use the web to sell access to their image collections. Despite significant security measures, images may be stolen and distributed, making it necessary to detect copyright violations. This demonstration paper describes a content-based system for large-scale automatic copyright enforcement. The paper briefly describes the image description, indexing and retrieval algorithm that lie at the heart of the system. It also describes our proposed demonstration, which is a realistic scenario of copyright violations of a large image collection.