Improving Automatic Video Retrieval with Semantic Concept Detection

Koskela, Markus; Sjöberg, Mats; Laaksonen, Jorma

doi:10.1007/978-3-642-02230-2_49

Cited by 6 publications

(4 citation statements)

References 12 publications

(16 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Concept-based video search has been found to be a promising direction for facilitating semantic video search and could outperform text-based or content-based video search [9]. Recently, a wealth of efforts have been devoted to various aspects of concept-based video search, including semantic concept detection, concept detector selection, search result fusion, and interactive search etc.…”

Section: Related Workmentioning

confidence: 98%

Utilizing related samples to learn complex queries in interactive concept-based video search

Yuan

Zha

Zhao

et al. 2010

Proceedings of the ACM International Conference on Image and Video Retrieval

View full text Add to dashboard Cite

One of the main challenges in interactive concept-based video search is the insufficient relevant sample problem, especially for queries with complex semantics. To address this problem, in this paper, we propose to utilize "related samples" to learn the complex queries. The "related samples" refer to those video segments that are irrelevant to the query but relevant to some of the related concepts of the query. Different from the relevant samples which may be rare, the related samples are usually sufficient and easy to find in the search result list. Specifically, we learn a detector for the query by simultaneously leveraging the related concept detectors, as well as users' feedbacks including relevant, irrelevant, and related samples. The query detector is then employed to predict the presence of the query in new video segments. As a result, new search results can be obtained according to the query presence. Furthermore, our approach is developed based on incremental learning technique. Thus, the query detector can be efficiently updated in each feedback iteration. We conduct experiments on two real-world video datasets: TRECVID 2008 and Youtube datasets. The experimental results demonstrate the effectiveness and efficiency of the proposed approach.

show abstract

Section: Related Workmentioning

confidence: 98%

Utilizing related samples to learn complex queries in interactive concept-based video search

Yuan

Zha

Zhao

et al. 2010

Proceedings of the ACM International Conference on Image and Video Retrieval

View full text Add to dashboard Cite

show abstract

“…This can be useful for concept detection and video search tasks [4], but organisation based on low-level features is not necessarily easily interpreted by humans, and such maps are not always useful for visualisation or browsing tasks. However, the map resulting from applying the SOM algorithm to semantic concept features can be expected to organise the data set semantically, placing conceptually similar objects nearby on the map surface.…”

Section: Related Workmentioning

confidence: 99%

“…On the other hand one has much more examples to use for training. To answer the second question posed above, we also perform the same experiments with two fast but weak classification methods: a simple linear classifier (logistic regression solver) and the PicSOM method [4], which is based on the Self-Organising Map (SOM) [5] algorithm. The PicSOM system in particular requires no per-concept training, the structure of the database needs to be learned only once, after which any new concept can be detected instantaneously by simply projecting the labelled examples onto the map.…”

Section: Introductionmentioning

confidence: 99%

Using semantic features to improve large-scale visual concept detection

Sjöberg

Laaksonen

2014

2014 12th International Workshop on Content-Based Multimedia Indexing (CBMI)

Self Cite

View full text Add to dashboard Cite

Currently there are many multimedia benchmarks and databases available with a predefined set of concepts for which detectors can be formed or are even already available. One can use these background concepts to form semantic concept vectors for each image or video in the database by concatenating the concept prediction outputs. In this paper we investigate the use of such semantic concept features for detecting novel concepts in two large-scale experiments: the TRECVID 2012 evaluation with 800 hours of video data, and MIRFLICKR with 1 million images. We show that the detection performance can improve significantly over using visual features only. In some applications, computationally expensive kernel classifiers cannot be used in the detection phase, and our experiments show a consistent significant improvement using fast linear classifiers when we replace visual features with the semantic concept feature. We also propose a Self-Organising Map-based method which affords fast trainingfree detection and intuitive visualisation properties.

show abstract

“…However, the approaches in this category su®er from the \semantic gap" issue [7] since the low-level features are insu±cient to describe the high-level semantic concepts. Therefore, the performance of a ranking algorithm based on the context and/or content information is usually limited.…”

Section: Introductionmentioning

confidence: 99%

Re-Ranking Algorithm for Multimedia Retrieval via Utilization of Inclusive and Exclusive Relationships Between Semantic Concepts

Chen

Shyu

2012

Int. J. Semantic Computing

View full text Add to dashboard Cite

The performance of ranking the retrieval results based on context and/or content information is usually limited because it depends heavily on the quality of the extracted information. Therefore, re-ranking approaches have been developed to improve the retrieval performance, but how to re-rank the retrieval results remains a challenge. In this paper, the inclusive and exclusive relationships between semantic concepts are utilized to re-rank the retrieval results. The inclusive relationship refers to the high co-occurrence relation between a target semantic concept and a reference semantic concept; while the exclusive relationship refers to the low or none co-occurrence relation. Moreover, multiple reference concepts provide more information from di®erent aspects than only one reference concept does. Multiple Correspondence Analysis (MCA) is applied in this paper to capture the correlation in terms of the impact weights between attributes and the co-occurrence classes, and the average value of all the impact weights is utilized to re-rank the retrieved results for the target semantic concept. Experimental results on the MediaMill Challenge problem and the TRECVID 2011 video collections demonstrate that our proposed framework can enhance the retrieval results for a number of target semantic concepts.

show abstract

Improving Automatic Video Retrieval with Semantic Concept Detection

Cited by 6 publications

References 12 publications

Utilizing related samples to learn complex queries in interactive concept-based video search

Utilizing related samples to learn complex queries in interactive concept-based video search

Using semantic features to improve large-scale visual concept detection

Re-Ranking Algorithm for Multimedia Retrieval via Utilization of Inclusive and Exclusive Relationships Between Semantic Concepts

Contact Info

Product

Resources

About