Main Page File List Related Pages
ITS_Graph3.java File Reference
File containing the ITS_Graph3 object used to create the different components of the Graphical User Interface.
Note that this version is compatible with the graph at format 2 (complete graph with nodes and edges), and with images and classfiles of sizeImxsizeIm pixels. Because this format is not optimized, still requires a lot of memory. Therfore, should be used with a little data set.
More...
Go to the source code of this file.
Compounds |
class | ITS_Graph3 |
Detailed Description
File containing the ITS_Graph3 object used to create the different components of the Graphical User Interface.
Note that this version is compatible with the graph at format 2 (complete graph with nodes and edges), and with images and classfiles of sizeImxsizeIm pixels. Because this format is not optimized, still requires a lot of memory. Therfore, should be used with a little data set.
This visual interface was developped in order to create a first tool to naviguate through the generated image time-series representations, and therfore to validate the results. As the JAVA programming was a completely new language for me and as not much time was available, the following developpement has been experimental, to avoid saying chaotic.
Nevertheless, this tool enabes the visualisation and the synchronized naviguation through the image time-series representation, the graph of dynamic cluster trajectories and the dynamic classfiles. Futhermore, if the interactive learning C++ program is launch in parallel, the current semantic labelling and the relevance feedback measurements learned by the system can be visualized by this JAVA GUI (The communication between the C and the JAVA program is done using text files).
The frame contains
- a) the image time-series sub-frame
- b) the graph sub-frame
- c) a TL classfile sub-frame
- d) a MT classfile sub-frame.
- e) a sub-frame for relevance feedbacks visualisation
- f) a sub-frame to naviguate through the current sub-graph a posteriori collection
- and g) a sub-frame for text values visualisation (sub-graph spatio-temporal location, sub-graph posterior probabilities, etc...).
User can interact with almost all these sub-frames :
-> In the sub-frame a) the user can naviguate in time through all the signal representations by pressing buttons. By clicking on a images location, he can naviguate in space and visualize corresponding MT classes, TL classifications, and graph representations
-> In sub-frame b) the user can visualize the dynamic cluster trajectories associated to the current MT classes (a maximum of 2 MT classes are visualisable at the same time in yellow and red). The mutual information is visualized in the same frame. The user can update all the other representations by dfining a time coordinate by simple click in the sub-frame. The complexity of the graph can also be fixed interactively, by adjusting the divergence treshold. Note that, consequently the TL classes associated to the MT classes are updated.
-> In sub-frame c) the user visualize the dynamic classifications. They are constituted of TL classes associted to the 2 different MT classes (commun TL classes are displayed in black, and in red and yellow for TL classes associted to only one of the 2 MT classes). Divergence treshold can be tuned to extend or concentrate these dynamic classification.
-> In sub-frame d) the MT classes are visualized. The user can select the MT classes by clicking on the image intersting locations or by pressing buttons. A mximum of 2 MT classses are visualized in this frame : they are visualized in yellow and red in correspondance with the cluster trajectories and the dynamic classifications.
-> Sub-frame e) is dedicated to visualize divergence histograms related to the current state of learning. No interaction exist in this frame
In sub-frame f), to naviguate through the current sub-graph a posteriori collection, press buttons are used to read an ./APostlist text file containing relevance feeback measurements and the sub-graphs a posteriori sotred collection.
-> Sub-frame g) is dedicated to the visualisation of text parameters. No interaction exist in this frame
PARAMETERS :
The input parameters are :
- the path of the images JPG quicklooks
- the path of the MT classfile
- the paths of the TL classfiles
- the path of the graph at format 2
-
Todo:
-
The interactive learning procedure has to be incorporated so that this interface could support the visualisation and naviguation through the representations, but also so that it could be able to drive the interactive learning alone (without the C++ program communicating with this applet using textfiles).
Moreover, the GUI has to be extend to support bigger images than sizeImxsizeIm pixels using visualization tools to naviguate in the full resolution images (zoom, quiklooks generator, etc..).
The graph visualisation could be enhanced : visualisation of all the graph attributes, enhancing the clarety of the visualisation (color, shape, etc...), multidimenionnal visualisation (visulaizing the 3D cluster trajectories), etc.
As this program is very experimental and very chaotic, DOCUMENTING IT WOULD NOT ENHANCE ITS CLARETY. Therefore, I would recommend, for future, to RE-CODE THIS APPLET FROM SCRATCH, tacking into account the above remarks.
-
Author:
-
P.Héas (IRIT / ENSAE / DLR / CNES)
-
Since:
-
January 2005
Definition in file ITS_Graph3.java.
Generated on Thu Feb 17 11:04:51 2005 for Graphical User Interface for mining image time-series by
1.2.14 written by Dimitri van Heesch,
© 1997-2002