Adaptive image retrieval using a Graph model for semantic feature integration
read more
Citations
Information retrieval
On social networks and collaborative recommendation
Categorising social tags to improve folksonomy-based recommendations
Interactive search in image retrieval: a survey
Generating Visual Summaries of Geographic Areas Using Community-Contributed Images
References
The anatomy of a large-scale hypertextual Web search engine
The Anatomy of a Large-Scale Hypertextual Web Search Engine.
Visual pattern recognition by moment invariants
Content-based image retrieval at the end of the early years
Image Processing: Analysis and Machine Vision
Related Papers (5)
Frequently Asked Questions (14)
Q2. What is the disadvantage of this method?
The disadvantages of this method is that it uses the original indices and similarity computations rather than the graph representation to determine feature weights.
Q3. What is the advantage of the inverted index?
Instead of having to compare N vectors given a particular query, the inverted index facilitates a fast computation of the relevant results.
Q4. What are the problems addressed in this paper?
The problems addressed in this paper are (a) how to capture and model personalised usage information to improve retrieval performance, and (b) how to integrate this information with other features (visual and textual) to model interdependencies between features.
Q5. How many images are better suited to the annotation graph?
In the image annotation graph of [12] a value of α = 0.65 was found to be better suited, which they could explain by a relationship to the estimated diameter of the graph.
Q6. What is the cosine similarity between a query vector and a document?
The cosine similarity between a query vector Q and a document Di is defined assim(Q,Di) = ∑Vj=0 wQ, j ×wi, j√∑Vj=0(wQ, j)2 ∑ V j=0(wi, j)2(12)where V is the total number of terms, and wQ, j is the weight of term j in the query.
Q7. How many iterations are used to get relevance feedback?
The setup of these runs is the following: for each task 200 queries consisting of 3 example images are issued to the system and relevance feedback is performed over a total of 20 relevance feedback iterations.
Q8. How is the effectiveness of the proposed approach evaluated?
Through systematic experimental results the effectiveness ofthe proposed approach is validated and learning strategies are investigated.
Q9. What are the links between nodes in an image-context graph?
The links between nodes represent: (1) image attributes (relations between images and their features); (2) intra-feature relations (feature similarities); and (3) semantic relations (peer information).
Q10. What is the way to determine the similarity between query nodes?
the graph structure could be used to determine the similarity between query nodes based on the three individual features.
Q11. What are the sources of information for the retrieval system?
In addition to the peer information, low-level visual features and textual annotations are further sources of information for the retrieval (and recommendation) system.
Q12. What is the common method of removing dangling pages?
To create a stochastic, irreducible matrix, Brin and Page suggested to eliminate dangling pages (pages with no outlinks) by linking them to all other pages in the Web [1].
Q13. What is the way to use the feature weights?
Instead of simply using the feature weights as a scaling factor for updating link weights, the authors can also envisage a more drastic weighting technique.
Q14. What is the optimal query vector for the i-th feature?
The optimal query vector ~qi (for the i-th feature) is calculated as the centroid of the P positive examples specified by the user.