UQMG @ TRECVid 2015: Instance Search
نویسندگان
چکیده
The UQMG group submits three runs for instance search at TRECVid 2015 [13]: all of them are automatic runs. Instead of adopting the traditional retrieval approach, e.g., Bag-of-Visual-Word (BoVW), our approach consists of three major steps: video decomposition, feature extraction and indexing. During decomposition, video segmentation is applied and various objects are extracted. Here a visual object is a minimal unit, and a video might consists of thousands of objects. Then we extract the visual feature of the object by using a convolutional neural network (ConvNet), which is a high-dimensional vector outputted by a fully connected layer of the network. Finally, the instance search problem is treated as finding the approximate nearest neighbors (ANN) of a given query in a large set of data points in high-dimensional space. Our best mAP is 0.114.
منابع مشابه
TRECVid 2012 Experiments at Dublin City University
Following previous participations in TRECVid, this year, the DCU-IAD team participated in four tasks of TRECVid 2012: Instance Search (INS), Interactive Known-Item Search (KIS), Multimedia Event Detection (MED) and Multimedia Event Recounting (MER).
متن کاملJOANNEUM RESEARCH and Vienna University of Technology at TRECVID 2010
We participated in two tasks: semantic indexing (SIN) and instance search (INS).
متن کاملTRECVID 2010 Experiments at Dublin City University
This year the DCU-CLARITY-iAD team participated in the both the instance search and interactive known-item search tasks of TRECVid 2010. For our instance search submission, we used classifiers to search for candidate objects in each keyframe. This was achieved by a coarse-to-fine search based on a hierarchical representation of the regions in the keyframes. Our results proved inconclusive, but ...
متن کاملThe University of Sheffield and Harbin Engineering University at TRECVID 2012 : Instance Search
This paper describes our contribution to instance search (INS) task for TRECVID 2012. We present four approaches for this task, (i) histograms of SIFT features as feature vectors and Bhatacharya distance for similarity detection (ii) feature vector is combination of SIFT features alone, while for matching we used a basic descriptor matching algorithm (iii) IR based approach using SIFT features ...
متن کاملARTEMIS-UBIMEDIA at TRECVid 2011: Instance Search
This paper describes the approach proposed by ARTEMISUBIMEDIA team at TRECVID 2011, Instance Search (INS) task. The method is based on a semi-global image representation relying on an over-segmentation of the keyframes. An aggregation mechanism was then applied in order to group a set of sub-regions into an object similar to the query, under a global similarity criterion.
متن کامل