I am doing research in content based image retrieval using neural networks. How can I implement an unsupervised method to retrieve the images?
Image retrieval looks like a supervised task. If you have an image query and want to find similar images, you can use a SOM network where each neuron represents some images then the winner neuron will return those similar images.
I have done some work in content-based image retrieval using multimodal keywords.
See the following papers:
Bayesian Framework for Automatic Image Annotation using Visual Keywords", Rajeev Agrawal, Changhua Wu,
William I. Grosky, Farshad Fotouhi, International Conference on Ubiquitous Computing and Multimedia Applications
(UCMA 2010).
"Searching an Appropriate Template Size for Multimodal Image Clustering", Rajeev Agrawal, William I. Grosky, Farshad
Fotouhi. International Conference on Multimedia Computing and Systems (ICMCS 2009).
"Diffusion Maps Based Image Clustering", Rajeev Agrawal, Chang-Hua Wu, William I. Grosky, Farshad Fotouhi ,
International Workshop On Research Issues in Digital Libraries, (IWRIDL 2006). ISBN No.: 1-59593-608-4
"Image Retrieval Using Multimodal Keywords", Rajeev Agrawal, William I. Grosky, Farshad Fotouhi, International
Symposium on Multimedia 2006: 817-822.
"Image Clustering Using Multimodal keywords", Rajeev Agrawal, William I. Grosky, Farshad Fotouhi, First International
Conference on Semantics And Digital Media Technology (SAMT 2006), Athens, Greece, Dec 6-8, 2006.
Hopfield neural networks? They are designed to provide a content addressable memory... and I guess they are considered non-supervised because you just apply the hebb rule to let them memorize.
Check out here:
https://en.wikipedia.org/wiki/Hopfield_net
and for a (really good) math explanation:
http://page.mi.fu-berlin.de/rojas/neural/chapter/K13.pdf
To use SOMs is adaquate BUT to take the winner neuron only not, as SOMs show a "jumping behaviour" , means you will have uncertain results. For that reason we developed the method of "Computing with Actvities" which takes the overall activity pattern of the SOM in to account. It furtheron the SOM achitecture is closed you will have a powerfull storage (16high number of neurons patterns) which will enable you to make a differentialy patern recognition.
thanx for all ur support and advise.suggest me to apply hopfield NN to any of content based image data base...
Jennifer Dy, Carla Brodley, A. C. Kak, Lynn Broderick, and Alex Aisen, "Unsupervised Feature Selection Applied to Content-Based Retrieval of Lung Images," IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 373-378, March 2003.
is the above paper has good credit...... to start
- Badges
- Science topic