Fabrizio Falchi

On scalability of the similarity search in the world of peers

Due to the increasing complexity of current digital data, similarity search has become a fundamental computational task in many applications. Unfortunately, its costs are still high and the linear scalability of single server implementations prevents from efficient searching in large data volumes. In this paper, we shortly describe four recent scalable distributed similarity search techniques and study their performance of executing queries on three different datasets. Though all the methods employ parallelism to speed up query execution, different advantages for different objectives have been identified by experiments. The reported results can be exploited for choosing the best implementations for specific applications. They can also be used for designing new and better indexing structures in the future.

Scalability comparison of Peer-to-Peer similarity search structures

Due to the increasing complexity of current digital data, similarity search has become a fundamental computational task in many applications. Unfortunately, its costs are still high and grow linearly on single server structures, which prevents them from efficient application on large data volumes. In this paper, we shortly describe four recent scalable distributed techniques for similarity search and study their performance in executing queries on three different datasets. Though all the methods employ parallelism to speed up query execution, different advantages for different objectives have been identified by experiments. The reported results would be helpful for choosing the best implementations for specific applications. They can also be used for designing new and better indexing structures in the future.

A metric cache for similarity search

Similarity search in metric spaces is a general paradigm that can be used in several application fields. It can also be effectively exploited in content-based image retrieval systems, which are shifting their target towards the Web-scale dimension. In this context, an important issue becomes the design of scalable solutions, which combine parallel and distributed architectures with caching at several levels. To this end, we investigate the design of a similarity cache that works in metric spaces. It is able to answer with exact and approximate results: even when an exact match is not present in cache, our cache may return an approximate result set with quality guarantees. By conducting tests on a collection of one million high-quality digital photos, we show that the proposed caching techniques can have a significant impact on performance, like caching on text queries has been proved effective for traditional Web search engines.

Caching content-based queries for robust and efficient image retrieval

In order to become an effective complement to traditional Web-scale text-based image retrieval solutions, content-based image retrieval must address scalability and efficiency issues. In this paper we investigate the possibility of caching the answers to content-based image retrieval queries in metric space, with the aim of reducing the average cost of query processing, and boosting the overall system throughput. Our proposal exploits the similarity between the query object and the cache content, and allows the cache to return approximate answers with acceptable quality guarantee even if the query processed has never been encountered in the past. Moreover, since popular images that are likely to be used as query have several near-duplicate versions, we show that our caching algorithm is robust, and does not suffer of cache pollution problems due to near-duplicate query objects. We report on very promising results obtained with a collection of one million high-quality digital photos. We show that it is worth pursuing caching strategies also in similarity search systems, since the proposed caching techniques can have a significant impact on performance, like caching on text queries has been proven effective for traditional Web search engines.

Deep learning for decentralized parking lot occupancy detection

We propose an effective CNN architecture for visual parking occupancy detection.The CNN architecture is small enough to run on smart cameras.The proposed solution performs and generalizes better than other SotA approaches.We provide a new training/validation dataset for parking occupancy detection. A smart camera is a vision system capable of extracting application-specific information from the captured images. The paper proposes a decentralized and efficient solution for visual parking lot occupancy detection based on a deep Convolutional Neural Network (CNN) specifically designed for smart cameras. This solution is compared with state-of-the-art approaches using two visual datasets: PKLot, already existing in literature, and CNRPark-EXT. The former is an existing dataset, that allowed us to exhaustively compare with previous works. The latter dataset has been created in the context of this research, accumulating data across various seasons of the year, to test our approach in particularly challenging situations, exhibiting occlusions, and diverse and difficult viewpoints. This dataset is public available to the scientific community and is another contribution of our research. Our experiments show that our solution outperforms and generalizes the best performing approaches on both datasets. The performance of our proposed CNN architecture on the parking lot occupancy detection task, is comparable to the well-known AlexNet, which is three orders of magnitude larger.

kNN based image classification relying on local feature similarity

In this paper, we propose a novel image classification approach, derived from the kNN classification strategy, that is particularly suited to be used when classifying images described by local features. Our proposal relies on the possibility of performing similarity search between image local features. With the use of local features generated over interest points, we revised the single label kNN classification approach to consider similarity between local features of the images in the training set rather than similarity between images, opening up new opportunities to investigate more efficient and effective strategies. We will see that classifying at the level of local features we can exploit global information contained in the training set, which cannot be used when classifying only at the level of entire images, as for instance the effect of local feature cleaning strategies. We perform several experiments by testing the proposed approach with different types of image local features in a touristic landmarks recognition task.

Automatic metadata extraction and indexing for reusing e-learning multimedia objects

In this paper we present the architecture of a Digital Library for enabling the reusing of audiovisual documents in an e-Learning context. The reuse of Learning Objects is based on automatically extracted descriptors carrying a semantic meaning for the professional that uses these Learning Objects to prepare new interactive multimedia lectures. The presented system is based on MILOS, a general purpose Multimedia Content Management System created to support design and effective implementation of digital library applications. MILOS supports the storage andcontent based retrieval of any multimedia documents whose descriptions are provided by using arbitrary metadata models represented in XML. The objective is to demonstrate the reuse of digital content, as video documents or Power Point presentations, by exploiting existing technologies for automatic extraction of metadata (OCR, speech recognition, cut detection, MPEG-7 visual descriptors, etc.). The search interface assists the user of the system in the retrieval the multimedia objects in the collection, by combining full-text retrieval on text extracted and metadata, and similarity search on the MPEG-7 visual descriptors.

Nearest neighbor search in metric spaces through Content-Addressable Networks

Most of the peer-to-peer search techniques proposed in the recent years have focused on the single-key retrieval. However, similarity search in metric spaces represents an important paradigm for content-based retrieval in many applications. In this paper we introduce an extension of the well-known Content-Addressable Network paradigm to support storage and retrieval of more generic metric space objects. In particular we address the problem of executing the nearest neighbors queries, and propose three different algorithms of query propagation. An extensive experimental study on real-life data sets explores the performance characteristics of the proposed algorithms by showing their advantages and disadvantages.

Combining local and global visual feature similarity using a text search engine

In this paper we propose a novel approach that allows processing image content based queries expressed as arbitrary combinations of local and global visual features, by using a single index realized as an inverted file. The index was implemented on top of the Lucene retrieval engine. This is particularly useful to allow people to efficiently and interactively check the quality of the retrieval result by exploiting combinations of various features when using content based retrieval systems.

Recognizing Landmarks Using Automated Classification Techniques: Evaluation of Various Visual Features

In this paper, the performance of several visual features is evaluated in automatically recognizing landmarks (monuments, statues, buildings, etc.) in pictures. A number of landmarks were selected for the test. Pictures taken from a test set were classified automatically trying to guess which landmark they contained. We evaluated both global and local features. As expected, local features performed better given their capability of being less affected to visual variations and given that landmarks are mainly static objects that generally also maintain static local features. Between the local features, SIFT outperformed SURF and ColorSIFT.