23.  Reference Manual
23.1    Input
23.2    Image Processing
23.3    Normalization
23.4    BioFilter
23.5    Neural Filter
23.6    Neural Net Filter
23.7    Batch Commands
23.8    BioFilter II
23.9    Neural Filter II
23.10   Neural Net II Filter
23.11   Segment-Locator
23.12   Other Objects In the Page
23.13   Examples


23.  Reference Manual

This manual is divided into the following sections:

1. Input
2. Image Processing
3. Normalization
4. BioFilter
5. Neural Filter
6. Neural Net Filter
7. Batch
8. BioFilter II
9. Neural Filter II
10. Neural Net Filter II
11. SegLocator
12. Miscellaneous
13. Examples
Each section is further divided into 3 parts: Filter Selection uses the Drop Down List. Menu Items use the menu bar. Parameters are set by using the “Parameter” button to open a new window.

23.1  Input

Key-Segment Button

Use the “Key-Segment” button to specify a key. For example, click the button, go to directory, "C:\images\", and then click "attrasoft.jpg". The selected image will be shown in the ImageFinder to indicate the specification is successful and the key-image is ready for training or retraining.
Search Dir Button
Use the “Search Dir” button to specify a search-directory. To select a search-directory, go to the directory; then click any file in the directory. The first image in the search-directory will be shown in the ImageFinder to indicate the specification is successful and the search-directory is active for retrieval.
File Input Button
Use the “File Input” button to specify a search-file.
Show Seg Button
Use the “Show Seg” button to display the selected training image.
ShowFile Button
Use the “ShowFile” button to display the selected search-file.
First, Next Button for Search Dir
Use the “First” button to display the first image in the search-directory. Use the “Next” button to display the next image in the search-directory.
First, Next Button for Search-File
Use the “First” button to display the first image in the search-file. Use the “Next” button to display the next image in the search-file.


23.2   Image Processing

Edge Filter Drop Down List

Use the “Edge Filter Drop Down List” to select an Edge Filter. The Edge Filter is an optional filter. The Edge Filters attempt to exaggerate the main feature(s) a user is looking for. The Edge Filters usually require a dark threshold filter. The Edge Filters extract and enhance edges & contours in an image by expressing intensity differences (gradients) between neighboring pixels as an intensity value. The basic variables are the differences between the top and bottom rows; the differences between the left and right columns; and the differences between the center point and its neighbors.

The batch codes for the Edge Filters are:

Code   Meaning
0         No Edge Filter
1         Sobel 1 (Prewitt)
2         Sobel 2 (Sobel)
3         Sobel 3
4         Sobel 4
5         Gradient
6         Gradient, 45°
7         Sobel 1, 45°
8         Sobel 1, - 45°
9         Laplacian 4
10        CD 11
11        FD 11
12        FD 9
13        FD 7
14        Laplacian 5
15        Laplacian 8
16        Laplacian 9
17        Laplacian 16
18        Laplacian17
 

Threshold Filter Drop Down List
Use the “Threshold Filter Drop Down List” button to set the Threshold Filter. The Threshold Filters attempt to suppress the background. The Threshold Filter is not optional; if you do not set the Threshold Filter, the default filter will be used.

Choose an Edge Filter and a Threshold Filter where the sample objects will stand out, otherwise change the filter. If you do not have a filter, a customized filter has to be built. DO NOT make too many things stand out, i.e. as long as the area of interest stands out, the rest should show as little as possible.

Once you make a selection, the objects in the training images are black and the background is white. You should make the black area as small as possible, as long as it covers the key-segment(s). Otherwise, switch to a different filter.
 

CleanUp Filter Drop Down List
Use the “CleanUp Filter Drop Down List” to select a Clean Up Filter. The CleanUp Filters will smooth the resulting image to reduce recognition error. The CleanUp Filter is an optional filter.
23.3   Normalization

Reduction Filter Drop Down List

Use the “Reduction Filter Drop Down List” to select a Reduction Filter. When reducing images, a scaling factor can be introduced easily. Although scaling symmetry can compensate for this scaling factor, the scaling symmetry is expensive. The Internal Reduction parameter is introduced to avoid unnecessary scaling.

There are several ways to reduce images:

  • Integer;
  • Real; or
  • All images are reduced by a same amount (Customized Version Only).
    • Integer Reduction
    Images are reduced by an integer factor to maximally fit 100x100 without distortion. For example, a 350x230 image will be reduced to 87x57.


    Real Reduction

    Images are reduced by a real number to maximally fit 100x100 without distortion. For example, a 350x230 image will be reduced to 100x65.


    All

    All training images and images in the search-directory are reduced by the same integer to fit 100x100 without distortion.


    Within each type of reduction, there are 3 more settings:

  • Avg: AOI (Area Of Interest) is half black and half white;
  • Max: AOI is mostly white, or the black areas are thin lines; or
  • Min: AOI is mostly black.
    • Reduction Filter Parameter / Segment-Cut Button
    Use the “Segment-Cut” button to shape the segment considered by the ImageFinder. The Segment-Cut parameter ranges from 0 to 12. This parameter deals with the edges of segments in the images. The larger this parameter is, the smaller the segment the ImageFinder will use. The possible settings of this parameter in the user interface are: 0, 1, 2, ..,and 12. To set the parameter, keep clicking the button.
    Reduction Filter Parameter/Size Cut Button
    Use the "Size-Cut" button to limit the dimensions of images to be searched. In some applications, the users only want to search images of certain dimensions and ignore other images.

    The dimension setting ranges from 0 to 9. To set the parameter, keep clicking the “Dimension” button; the setting will switch from one to the next each time you click the button.

    If the setting is 0, this parameter will be ignored. If the parameter is 1, then the longest edge of the image to be considered must be at least 100, but less than 199. If the parameter is 2, then the longest edge of the image to be considered must be at least 200, but less than 299, …

    Reduction Filter Parameter / Border Cut
    Use the “Border-Cut” button to ignore the sections of the image near the borders. The Border-Cut parameter ranges from 0 (no cut) to 9 (18% border cut). The possible settings in the user interface are: 0, 1, 2, ..,and 9. Assume an image is (0,0; 1,1); setting Border-Cut to 1 means the ImageFinder will look at the section (0.02, 0.02; 0.98, 0.98); setting Border-Cut to 2 means the ImageFinder will look at the section (0.04, 0.04; 0.96, 0.96); … . To set the parameter, keep clicking the button.
    Reduction Filter Parameter / Look-At Area
    The “Look-At Area” is the area the ImageFinder will use. A 100 x 100 window specifies a whole image. In the Integer-Reduction, the actual area can be less than 100x100.  The Look-At Area is specified by 4 numbers:

     (x, y, w, h)

    (x, y) are the coordinates of the upper-left corner and (w, h) are the width and height of the Look-At Window.

    To use this Look-At Window, enter values for (x, y, w, h) in the four textboxes. Note that the image display area in the ImageFinder is 300x300, therefore, the training segment is specified within a 300x300 area. The Look-At Window is 100x100. The default value is (0,0,0,0), meaning the Look-At area setting is ignored.


    23.4   BioFilter

    BioFilter Drop Down List

    Use the “BioFilter Drop Down List” to select a BioFilter. The main role of the BioFilter is (1) to make an assessment of data via Unsupervised Learning; and (2) to eliminate 80% of the mismatches.
    “BioFilter/Scan Images  - Directory Input” Menu Item
    Use “BioFilter/Scan Images  - Directory Input” menu item to convert images into records. The BioFilter is one of two filters in Feature Space image recognition (Image recognition is divided into Feature Space recognition and Input Space recognition. Feature Space recognition operates on signatures of images.) To convert images into templates:
  • Click “Search Dir” button to specify the search-directory.
  • Click menu item “BioFilter/Scan Images  - Directory Input” to convert images to records. You should see the ImageFinder scan through the images at this point.
    • “BioFilter/Scan Images  - File Input” Menu Item
    Use “BioFilter/Scan Images  - File Input” menu item to convert images into records. The BioFilter is one of two filters in Feature Space image recognition (Image recognition is divided into Feature Space recognition and Input Space recognition. Feature Space recognition operates on signatures of images.) To convert images into templates:
  • Click the “Input File” button to specify the search-file.
  • Click menu item “BioFilter/Scan Images  - File Input” to convert images to records. You should see the ImageFinder scan through the images at this point.
    • “BioFilter\Train (match.txt required)”  Menu Item
    Use “BioFilter\Train (match.txt required)” menu item to train the BioFilter. Training uses the data collected in advance to teach the BioFilter how to match. Training requires two files, a1.txt and match.txt:
  • A1.txt is the record file, which contains many records. Each image is converted into a record. A record represents features of an image in a Feature Space.
  • Match.txt is a list of matching pairs. This file will teach the ImageFinder who will match with whom.
    • “BioFilter/BioFilter 1:N Match (Untrained)” Menu Item
    Use “BioFilter/BioFilter 1:N Match (Untrained)” menu item to make an untrained 1:N Matching. 1:N Matching compares one key image with the images in a search-directory or search-file; the key image is specified in the “Key Segment” textbox or selected by the “Key Segment” button. 1:N Matching requires the images in the search-directory or search-file being converted into templates in advance. To make an 1:N Matching:
  • Click “Key Segment” button, and select an image;
  • Click “BioFilter/BioFilter 1:N Match (Untrained)”.
    • The result is in file, b1.txt, which will be opened at the end of computation.
    “BioFilter/BioFilter N:N Match (Untrained)” Menu Item
    Use “BioFilter/BioFilter N:N Match (Untrained)” menu item to make an untrained N:N Matching. N: N Matching compares each image, specified in the search-directory or search-file, with every image in the search-directory or search-file. N:N Matching requires the images in the search-directory or search-file being converted into templates in advance. The result is in file, b1.txt, which will be opened at the end of computation.
    “BioFilter/BioFilter 1:N Match” Menu Item
    Use “BioFilter/BioFilter 1:N Match” menu item to make a 1:N Matching. 1:N Matching compares one key image with the images in a search-directory or search- file; the key image is specified in the “Key Segment” textbox or selected by the “Key Segment” button. 1:N Matching requires (1) the images in the search-directory or search-file being converted into templates in advance; and (2) the BioFilter being trained. To make an 1:N Matching:
  • Click “Key Segment” button, and select an image;
  • Click “BioFilter/BioFilter 1:N Match”.
    • The result is in file, b1.txt, which will be opened at the end of computation.
    “BioFilter/BioFilter N:N Match” Menu Item
    Use “BioFilter/BioFilter N:N Match” menu item to make a N:N Matching. N: N Matching compares each image, specified in the search- directory or search-file, with every image in the search-directory or search-file. N:N Matching requires (1) the images in the search- directory or search-file being converted into templates in advance; and (2) the BioFilter being trained. The result is in file, b1.txt, which will be opened at the end of computation.
    “BioFilter/BioFilter Results” Menu Item
    Use “BioFilter/BioFilter Results” menu item to open b1.txt, the file containing the last matching result.
    “BioFilter/Check (b1_matchlist.txt required)” Menu Item
    Use “BioFilter/Check (b1_matchlist.txt required)” menu item to check the matching results. If this is a test run (i.e., you know the correct answers), you can see the matching results in seconds. To test the results in b1.txt, you must prepare B1_matchlist.txt file, which indicates the matching pairs. Once b1.txt and B1_matchlist.txt are prepared, the command will let you know your matching results in seconds.
    “BioFilter/Option” Menu Item
    Use “BioFilter/Option” menu item to open a sub-menu, which allows you to store the templates files into b2.txt, b3.txt, b4.txt, and b5.txt.
    “BioFilter/Option/Scan Images  - Directory Input (a2.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - Directory Input” menu item but saves the results to a2.txt.
    “BioFilter/Option/Scan Images  - Directory Input (a3.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - Directory Input” menu item but saves the results to a3.txt.
    “BioFilter/Option/Scan Images  - Directory Input (a4.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - Directory Input” menu item but saves the results to a4.txt.
    “BioFilter/Option/Scan Images  - Directory Input (a5.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - Directory Input” menu item but saves the results to a5.txt.
    “BioFilter/Option/Scan Images  - File Input (a2.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - File Input” menu item but saves the results to a2.txt.
    “BioFilter/Option/Scan Images  - File Input (a3.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - File Input” menu item but saves the results to a3.txt.
    “BioFilter/Option/Scan Images  - File Input (a4.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - File Input” menu item but saves the results to a4.txt.
    “BioFilter/Option/Scan Images  - File Input (a5.txt)” Menu Item
    Similar to “BioFilter/Scan Images  - File Input” menu item but saves the results to a5.txt.
    BioFilter Parameter/Bio-Filter Scale TextBox
    Use Bio-Filter Scale textbox to control the amount of output. This parameter ranges from 0 to 100. The larger this number is, the more matches you will get. To set this parameter, enter a number between 0 and 100 to the text box.
    BioFilter Parameter/Bio-Filter Mode Button
    Use “Bio-Filter Mode” button to determine whether later filters will use this filter. Since BioFilter is one of the recognition filters, you may or may not use this filter. This parameter decides whether the BioFilter will be used.

    This parameter has three values:

    Untrained
    Trained
    Bypass
    where
  • The “Untrained” setting will do an image matching without training.
  • The “Trained” setting requires the BioFilter be trained first.
  • The “Bypass” setting will by pass this filter.
    • To set the parameter, keep clicking the button; the setting will switch from one to the next each time you click this button.
    BioFilter Parameter/Output Button
    Use “Output” button to decide whether you want to display the score or not. To determine if one image "matches" another image, they must be compared using a unique algorithm. Generally, the result of this comparison is a "score", indicating the degree to which a match exists. This score is then compared to a pre-set threshold to determine whether or not to declare a match.

    The Output parameter has two settings:

    No Scores
    Scores

    If the BioFilter is an intermediate step, this score will not show up in the output file. The “No Scores” setting (default setting) will not show the scores in the Output file. If the BioFilter is the only filter used in the matching, then you can show the score in the Output file by selecting the “Scores” setting. To set the parameter, keep clicking the button; the setting will switch from one to the next each time you click the “Blurring” button.

    BioFilter Parameter/BioFilter Threshold TextBox
    Use Threshold TextBox to set the BioFilter threshold. The result of image comparison is a "score", indicating the degree to which a match exists. This score is then compared to a pre-set Threshold to determine whether or not to declare a match. This parameter sets the threshold.

    To decide what threshold to use, you should make a test run first and look at the scores. Matching images have higher scores; unmatched images have lower scores. Select a threshold to separate these two groups. There will be a few images in the middle, representing both groups. Under these circumstances, the threshold selection depends on your applications. To set the Threshold parameter, enter a number into the Threshold text box.

    23.5   Neural Filter

    Neural Filter Drop Down List

    Use the “Neural Filter Drop Down List” to select a Neural Filter. The Neural Filter is the main matching filter for the Feature Space.
    “NeuralFilter\Train (match.txt required)” Menu Item
    Use “NeuralFilter\Train (match.txt required)” menu item to train the Neural Filter. Training uses the data collected in advance to teach the Neural Filter how to match. Training requires two files, a1.txt and match.txt:
  • A1.txt is the record file, which contains many records. Each image is converted into a record. A record represents features of an image in a Feature Space.
  • Match.txt is a list of matching pairs. This file will teach the ImageFinder who will match with whom.
    • “NeuralFilter/NeuralFilter 1:N Match” Menu Item
    Use “NeuralFilter/NeuralFilter 1:N Match” menu item to make a 1:N Matching. 1:N Matching compares one key image with the images in a search-directory or search-file; the key image is specified in the “Key Segment” textbox or selected by the “Key Segment” button. 1:N Matching requires (1) the images in the search-directory or search-file being converted into templates in advance; and (2) the NeuralFilter being trained. To make an 1:N Matching:
  • Click “Key Segment” button, and select an image;
  • Click “NeuralFilter/NeuralFilter 1:N Match”.
    • The result is in file, b1.txt, which will be opened at the end of computation.
    “NeuralFilter/NeuralFilter N:N Match” Menu Item, and
    “NeuralFilter/Query Set + Target Set/a1 + a1 ==> b1” Menu Item
    Use “NeuralFilter/NeuralFilter N:N Match” menu item to make an untrained N:N Matching. N:N Matching compares each image, specified in the search-directory or search-file, with every image in the search-directory or search- file. N:N Matching requires (1) the images in the search-directory or search-file being converted into templates in advance; and (2) the NeuralFilter being trained. The result is in file, b1.txt, which will be opened at the end of computation.
    “NeuralFilter/NeuralFilter N:(N-1) Match” Menu Item
    Use “NeuralFilter/NeuralFilter N:(N-1) Match” menu item to make an N: (N-1) Matching. Let a and b be two images; N:N Matching is {aa, ab, ba, bb} and the N : (N-1) Matching is {ab}. The N: N Matching has N * N comparisons; and the N : (N-1) Matching has N * (N-1 )/2 comparisons.  The purpose of N : (N-1) Matching is to reduce the number of comparisons. N:(N-1) Matching requires (1) the images in the search-directory or search- file being converted into templates in advance; and (2) the NeuralFilter being trained. The result is in file, b1.txt, which will be opened at the end of computation.
    “NeuralFilter/NeuralFilter Results” Menu Item
    Use “NeuralFilter/NeuralFilter Results” menu item to open b1.txt, the file containing the last matching result.
    “NeuralFilter/Check (b1_matchlist.txt required)” Menu Item
    Use “NeuralFilter/Check (b1_matchlist.txt required)” menu item to check the matching results. If this is a test run (i.e., you know the correct answers), you can see the matching results in seconds. To test the results in b1.txt, you must prepare B1_matchlist.txt file, which indicates the matching pairs. Once b1.txt and B1_matchlist.txt are prepared, the command will let you know your matching results in seconds.
    “NeuralFilter/Query Set + Target Set” Menu Item
    Use “NeuralFilter/Query Set + Target Set” menu item to open a sub-menu. These menu items allow you to compare images in file with images in another file.
    “NeuralFilter/ Query Set + Target Set/a1 + a2 ==> b2” Menu Item
    Similar to “NeuralFilter/NeuralFilter N:N Match” menu item, except that   “NeuralFilter/NeuralFilter N:N Match” menu item makes the following matching: a1 + a1 ==> b1 (images in a1.txt matches against images in a1.txt and the results are in b1.txt), this command makes the following match: a1 + a2 ==> b2 (images in a1.txt matches against images in a2.txt and the results are in b2.txt).
    “NeuralFilter/ Query Set + Target Set/a1 + a3 ==> b3” Menu Item
    Similar to “NeuralFilter/NeuralFilter N:N Match” menu item, except that   “NeuralFilter/NeuralFilter N:N Match” menu item makes the following matching: a1 + a1 ==> b1 (images in a1.txt matches against images in a1.txt and the results are in b1.txt), this command makes the following match: a1 + a3 ==> b3 (images in a1.txt matches against images in a3.txt and the results are in b3.txt).
    “NeuralFilter/ Query Set + Target Set/a1 + a4 ==> b4” Menu Item
    Similar to “NeuralFilter/NeuralFilter N:N Match” menu item, except that   “NeuralFilter/NeuralFilter N:N Match” menu item makes the following matching: a1 + a1 ==> b1(images in a1.txt matches against images in a1.txt and the results are in b1.txt), this command makes the following match: a1 + a4 ==> b4 (images in a1.txt matches against images in a4.txt and the results are in b4.txt).
    “NeuralFilter/ Query Set + Target Set/a1 + a5 ==> b5” Menu Item
    Similar to “NeuralFilter/NeuralFilter N:N Match” menu item, except that   “NeuralFilter/NeuralFilter N:N Match” menu item makes the following matching: a1 + a1 ==> b1(images in a1.txt matches against images in a1.txt and the results are in b1.txt), this command makes the following match: a1 + a5 ==> b5 (images in a1.txt matches against images in a5.txt and the results are in b5.txt).
    NeuralFilter Parameter/Neural-Filter Scale TextBox
    Use Neural-Filter Scale textbox to control the amount of output. This parameter ranges from 0 to 100. The larger this number is, the more matches you will get. To set this parameter, enter a number between 0 and 100 to the text box.
    NeuralFilter Parameter/Neural-Filter Mode Button
    Use the “Neural-Filter Mode” button to determine whether this filter will be used by later filters. Since NeuralFilter is one of the recognition filters, you may or may not use this filter. This parameter decides whether the NeuralFilter will be used.

    This parameter has two values:
     

    Trained
    Bypass
    where:
  • The “Trained” setting requires the NeuralFilter being used by later filters.
  • The “Bypass” setting will by-pass this filter.
    • To set the parameter, keep clicking the button; the setting will switch from one to the next each time you click this button.
    NeuralFilter Parameter/Neural-Filter Opening Button
    Use the “Neural-Filter Opening” button to control the “Openness” of this filter. This parameter has the following settings: A large opening will allow more matches in the results.
    NeuralFilter Parameter/Output Button
    Use the “Output” button to decide whether you want to display the score or not. To determine if one image "matches" another image, they must be compared using a unique algorithm. Generally, the result of this comparison is a "score", indicating the degree to which a match exists. This score is then compared to a pre-set threshold to determine whether or not to declare a match.

    The Output parameter has two settings:

    No Scores
    Scores

    If the NeuralFilter is an intermediate step, this score will not show up in the output file. The “No Scores” setting (default setting) will not show the scores in the output file. If the NeuralFilter is the only filter used in the matching, then you can show the score in the output file by selecting the “Scores” setting. To set the parameter, keep clicking the button; the setting will switch from one to the next each time you click the “Blurring” button.

    NeuralFilter Parameter/NeuralFilter Threshold TextBox
    Use Threshold TextBox to set the NeuralFilter threshold. The result of image comparison is a "score", indicating the degree to which a match exists. This score is then compared to a pre-set Threshold to determine whether or not to declare a match. This parameter sets the threshold.

    To decide what threshold to use, you should make a test run first and look at the scores. Matching images have higher scores; unmatched images have lower scores. Select a threshold to separate these two groups. There will be a few images in the middle, representing both groups. Under these circumstances, the threshold selection depends on your application. To set the Threshold parameter, enter a number into the Threshold text box.


    23.6   Neural Net Filter

    Neural Net Filter Drop Down List

    Use the “Neural Net Filter Drop Down List” to set the Neural Net Filter. The default filter is 100x100. All images are scaled down by an integer amount. For example, 640x480 will be scaled down 7 times to 91x68. You need to understand this; so when you translate, or rotate an image, you will make sure no additional scaling factors are introduced.

    The search speed crucially depends on the Neural Net Filter. For example, if the 50x50 filter is used, then the underlying neural net size is reduced by a factor of 4; and the neural computation speed will be increased by a factor of 16.

    The filters available are:

  • 100x100
  • 90x90
  • 80x80
  • 70x70
  • 60x60
  • 50x50
    • Let the speed of 100x100 representation be a base, then the overall speed for:
  • 90x90 representation is 1 times faster;
  • 80x80 representation is 1.6 times faster;
  • 70x70 representation is 2.7 times faster;
  • 60x60 representation is 5 times faster; and
  • 50x50 representation is 10 times faster.
    • NeuralNet/NeuralNet Train Menu Item
    Use the “NeuralNet Train” menu item to train the software to learn the active key (what image to look for). These commands will first delete all old training and start to train the software from the beginning. After clicking a button, wait 1 second. If everything is O.K., a message "Training End!" will be printed in the Status Text Area.
    NeuralNet/NeuralNet Retrain Menu Item
    Use the “Retrain” menu item to retrain the software to learn additional keys. If several keys are used for training, the first learning uses training and all the subsequent learning uses retraining. After clicking the button, wait 1 second. If everything is O.K., a message "Retraining End!" will be printed in the Status Text Area.
    NeuralNet/1:N Search Menu Item
    Use the "1:N Search" menu item to retrieve images in the search-directory. You must train the software first before searching. After clicking the button, wait until a web page is opened. If everything is O.K., a message "Retrieval End!" will be printed in the text area, then a web page with a list of retrieved images will be opened. Whenever you click a file name in the opened web page, the retrieved image will be shown. You might need to click the "Refresh" button.

    Note that, next to each image, an integer, called similarity, is printed. The larger the number is, the more similarity between the training image(s) and the retrieved images.

    NeuralNet/N:N Search Menu Item
    Use the "N:N Search" menu item to match each image in the search-directory against all other images in the directory. After clicking the button, wait until a web page is opened. If everything is O.K., a message "Retrieval End!" will be printed in the text area, then a web page with a list of retrieved images will be opened. Whenever you click a file name in the opened web page, the retrieved image will be shown. You might need to click the "Refresh" button. In the output file, each image in the search-directory has a block; the first line in a block is the input and the rest of the lines in a block is the output.

    Note that, next to each image, an integer, called similarity, is printed. The larger the number is, the more similarity between the training image(s) and the retrieved images.

    NeuralNet/Sort Menu Item
    Use the "Sort" menu item to sort the retrieved 1:N Matching results according to their weights. You might need to click the "Refresh" button. For N:N Matching, only the last block is sorted.
    NeuralNet/Long-Search Directory Menu Item
    Use the “Long-Search” Directory menu item to specify a search-directory. To select a search-directory, go to the directory; then click any file in the directory. The first image in the search-directory will be shown in the ImageFinder to indicate the specification is successful and the search-directory is active for retrieval.
    NeuralNet/1:N Long-Search Menu Item
    Use the "1:N Long-Search" menu item to retrieve images in the sub-directories of the search-directory. You must train the software first before searching. After clicking the button, wait until a web page is opened. If everything is O.K., a message "Retrieval End!" will be printed in the text area, then a web page with a list of retrieved images will be opened. Whenever you click a file name in the opened web page, the retrieved image will be shown. You might need to click the "Refresh" button.

    Note that, next to each image, an integer, called similarity, is printed. The larger the number is, the more similarity between the training image(s) and the retrieved images.

    NeuralNet/Long-Sort Menu Item
    Use the "Long-Sort" menu item to sort the retrieved Long-Search results according to their weights. You might need to click the "Refresh" button.
    NeuralNet/1:N File-Search Menu Item
    Use the "1:N File Search" menu item to retrieve images in the search-file. You must train the software first before searching. After clicking the button, wait until a web page is opened. If everything is O.K., a message "Retrieval End!" will be printed in the text area, then a web page with a list of retrieved images will be opened. Whenever you click a file name in the opened web page, the retrieved image will be shown. You might need to click the "Refresh" button.

    Note that, next to each image, an integer, called similarity, is printed. The larger the number is, the more similarity between the training image(s) and the retrieved images.

    NeuralNet/N:N File Search Menu Item
    Use the "N:N File Search" menu item to match each image in the search-file against all other images in the file. After clicking the button, wait until a web page is opened. If everything is O.K., a message "Retrieval End!" will be printed in the text area, then a web page with a list of retrieved images will be opened. In the output file, each image in the search-file has a block; the first line in a block is the input and the rest of the lines in a block is the output.

    Note that, next to each image, an integer, called similarity, is printed. The larger the number is, the more similarity between the training image(s) and the retrieved images.

    NeuralNet/File Sort Menu Item
    Use the "File Sort" menu item to sort the retrieved 1:N file matching results according to their weights. You might need to click the "Refresh" button. For N:N Matching, only the last block is sorted.
    NeuralNet/NeuralNet Results Menu Item
    Use “NeuralNet Results” menu item to open the last matching result.
    NeuralNet/NeuralNet Advisor Menu Item
    Use the “NeuralNet Advisor” menu item to get the values of the following parameters: Segment-Cut, Blurring, Internal Weight Cut, and Sensitivity. The ImageFinder has a ‘Parameter Advisor” to help you to locate the general range of these parameters. To use the Advisor, you must already know how to search.

    Procedure:

    1. Assume a sample image is sample.jpg. Put the sample image and all matched images in a directory, say, c:\image\.
    2. Enter the sample image, sample.jpg, and the search-directory, c:\image\ into the ImageFinder.
    3. Set the parameters rather large to make sure all the matched images are retrieved by the ImageFinder, say:
    Segment Cut = 0
    Blurring = 50
    Shape Cut = 90
    Internal Weight Cut = 90
    Sensitivity = 90
    4. Click the Advisor Menu Item.
    5. Click the “Get Parameter” button in the Advisor window; you will see the recommended parameter settings.
    The recommended values represent an average, so it might not fit some of your images, but it gives you a basic idea of the general intervals of these parameters.
    NeuralNet/Parameter/Symmetry
    Use the "Symmetry" button to set the symmetry. The symmetry settings are:
  • No symmetry;
  • Translation symmetry;
  • Rotation symmetry;
  • Scaling symmetry; and
  • Rotation and Scaling symmetry.
    • The default setting is “Tran”, the Translation symmetry. To set the symmetry, keep clicking the button; the setting will switch from one to the next each time you click the button.
    NeuralNet/Parameter/Translation Type Button
    Use “Translation Type” button to select the accuracy of the translation symmetry.  The Translation Type settings (and their codes) are:
  • Most Accurate (0);
  • Accurate (1); and
  • Least (2).
    • To set the Translation Type, keep clicking the “T Type” button; the setting will switch from one to the next each time you click the button. The default setting is 0, the most accurate setting.
    NeuralNet/Parameter/Scaling Type Button
    Use “Scaling Type” button to select the accuracy of the scaling symmetry.  The Scaling Type settings (and their codes) are:
  • Least Accurate (0);
  • Accurate (1);
  • Accurate (2); and
  • Most Accurate (3).
    • To set the Scaling Type, keep clicking the “S Type” button; the setting will switch from one to the next each time you click the button. The default setting is 0, the least accurate setting.
    NeuralNet/Parameter/Rotation Type Button
    Use the R Type (Rotation Type) buttons to set the Rotation Types. The settings are:
  • 360° rotation (0);
  • -5° to 5° rotation (1);
  • -10° to 10° rotation (2);
  • 360° rotation, accurate (3);
  • 360° rotation, more accurate (4);
  • 360° rotation, most accurate (5).
    • Other settings can be ordered in a Customized Version.

    To set the Rotation Type, keep clicking “R Type” button; the setting will switch from one to the next each time you click the button. The default setting is 360° rotation (0).

    NeuralNet/Parameter/Training Segment Text Boxes
    Use Training Segment Text Boxes to set the training segment. To enter a segment, first select an image. The selected image will be shown in the ImageFinder, 300 by 300 in size, to indicate the specification is successful. The key-segment is specified by 4 integers:  the upper-left corner (x, y) and the length and height (w, h) of the segment. Once the segment specification is successful, a black box will cover the selected area. If the selected area is not what you want, just re-select the area again.
    NeuralNet/Parameter/Blurring Text Box
    Use Blurring Text Box to control the amount of output. "0%"-Blurring means the exact match. When the "Blurring" is increased, you will get more and more similar images. As the Blurring goes higher, the speed will be slower. The Blurring settings range from 0 – 50. To set the Blurring, enter a number between 0 and 50 to the text box. The default setting is 5%.
    NeuralNet/Parameter/Sensitivity Text Box
    Use Sensitivity Text Box to adjust search segment size. The Sensitivity parameter ranges from 0 (least sensitive) to 100 (most sensitive).
  • To search small segment(s), use a high sensitivity search.
  • To search large segment(s), use low sensitivity search.
  • The higher the parameter is set, the more results you will get.
    • To set the Sensitivity, enter a number between 0 and 100 the text box. The default setting is 100.
    NeuralNet/Parameter/External Weight Cut (ExternalCut) Text Box
    Use the External Cut Text Box to eliminate those retrieved images with the weights below the External Cut. This parameter is also called Threshold. To set the External Cut, enter a number to the text box. The default setting is 0. In general, it is better to give no answer than a wrong answer. Assume you are searching images and all similar images have weights ranging from 1,000 to 10,000. It is possible that some other images will pop up with weights ranging from 10 to 100. To eliminate these images, you can set the External Cut to 1,000.
    NeuralNet/Parameter/Internal Weight Cut (Internal Cut) Text Box
    The Internal Cut plays the similar role as the External Cut. There are two differences between these two cuts:
  • The Internal Cut ranges from 0 to 100; and the ExternalCut can be any number;
  • The Internal Cut stops the images from coming out; whereas, the External Cut can bring the eliminated images back if you set the External Cut to 0. You might need to see the eliminated images sometimes for the purpose of adjusting parameters.
    • To set the Internal Cut, enter a number between 0 and 100 to the text box. The default setting is 100.
    NeuralNet/Parameter/Segment Size Button
    Use the “Segment Size” button to select segment size. The default setting is "L Segment".
  • To search large segments, use "L Segment" (Large Segment). For example, if a sample segment is one quarter of the sample image, it is a large segment
  • To search small segments, use "S Segment" (Small Segment). If the segment is 1/25 of the sample image, it is a small segment.
    • To set the segment size, keep clicking the Size button; the setting will switch from one to the next each time you click the button.

    Currently, "S Segment" only supports translation symmetry. If you need rotation and scaling symmetry, please use "L Segment". Additional symmetry can be added very quickly in a Customized Version.

    NeuralNet/Parameter/Image Type Button
    There are BW and Color images. For each of them, there are “sum-search”, “maximum-search”, and “average-search”. This generates 6 image types:
  • Bi-level 1 (0)
  • Bi-level 2 (1)
  • Bi-level 3 (2)
  • Color 1 (3)
  • Color 2 (4)
  • Color 3 (5)
    • "Bi-level 1” is like an integration of a function f(x); "Bi-level 2” is like a maximum value of f(x); and "Bi-level 3” is the average of the above two.

    "Bi-level 1" search will produce a higher weight than a "Bi-level 2" search. "Bi-level 3" search is in the middle. Similarly, a "Color 1" search will produce a higher weight than a "Color 2" search. "Color 3" is in the middle.

    To set the image type, keep clicking the “Image Type” button; the setting will switch from one to the next each time you click the “Image Type” button.

    NeuralNet/Parameter/File Display Type Button
    Use “File Display Type” button to set the output file type. The options are text file and html file.
    NeuralNet/Parameter/AutoSegment Button
    Use “AutoSegment” button to select a training segment automatically. The options are:
    Manual Segment
    AutoSegment 10
    AutoSegment 20
    AutoSegment 30
    “Manual Segment” setting requires you to select a training segment using the 4 Training Segment Text Boxes introduced earlier in this section. To select a segment by computer, do not use “Manual Segment” setting. To set this parameter, keep clicking the button; the setting will switch from one to the next each time you click the button.

    The “AutoSegment 10” setting will select a larger segment than “AutoSegment 20” setting, which in turn, will select a larger segment than the “AutoSegment 30” setting.

    NeuralNet/Parameter/Mode Button
    Use Neural Net Mode button to determine whether this filter will be used by later filters. Since the Neural Net filter is one of the recognition filters, you may or may not use this filter. This parameter decides whether the BioFilter will be used.

    This parameter has two values:

    Trained
    Bypass

    The “Trained” setting requires that later filters use the NeuralNet filter. The “Bypass” setting will by-pass this filter.

    To set the parameter, keep clicking the button; the setting will switch from one to the next each time you click this button.


    23.7   Batch Commands

    Batch/Set Execution Code Menu Item

    Use “Set Execution Code” menu item to set the Set Execution Code. There are many commands in the ImageFinder. Each command has an integer for identification. This integer is called Batch Execution Code. This number is used before you save your batch code.
    Batch/Save Menu Item
    Use the "Save" menu item to save the last ImageFinder setting in batch code.  The batch code is saved to a file, abm60.txt.
    Batch/Save 2 Menu Item
    Use the "Save 2" menu item to save the last ImageFinder setting in batch code.  The batch code is saved to a file, abm60_2.txt.
    Batch/Save 3 Menu Item
    Use the "Save 3" menu item to save the last ImageFinder setting in batch code.  The batch code is saved to a file, abm60_3.txt.
    Batch/Save 4 Menu Item
    Use the "Save 4" menu item to save the last ImageFinder setting in batch code.  The batch code is saved to a file, abm60_4.txt.
    Batch/Save 5 Menu Item
    Use the "Save 5" menu item to save the last ImageFinder setting in batch code.  The batch code is saved to a file, abm60_5.txt.
    Batch/Open Menu Item
    Use the "Open" menu item to open the batch code file saved by the Batch/Save command.
    Batch/Open 2 Menu Item
    Use the "Open 2" menu item to open the batch code file saved by the “Batch/Save 2” command.
    Batch/Open 3 Menu Item
    Use the "Open 3" menu item to open the batch code file saved by the “Batch/Save 4” command.
    Batch/Open 4 Menu Item
    Use the "Open 4" menu item to open the batch code file saved by the “Batch/Save 4” command.
    Batch/Open 5 Menu Item
    Use the "Open 5" menu item to open the batch code file saved by the “Batch/Save 5” command.
    Batch/Notes
    Click the “Notes” menu item to create an online note so you can remember which batch code is for which problem.
    Batch/Run
    Use the "Run" menu item to execute the batch code in the display area.
    Batch/Load
    Use the "Load" menu item to load the parameters specified by the batch code in the display area. The load command will not execute the batch code.


    23.8   BioFilter II

    BioFilter II Drop Down List

    Use the “BioFilter II Drop Down List” to select a BioFilter-II. BioFilter I matches the whole image, while BioFilter-II matches a part of an image. The main role of the BioFilter-II is (1) a make an assessment of data via unsupervised learning; and (2) to eliminate 80% of the mismatches.
    “BioFilter 2/Scan Images  - Directory Input” Menu Item
    Use “BioFilter 2/Scan Images  - Directory Input” menu item to convert images into records. The BioFilter-II is one of two filters operating on image segments in feature space image recognition (Image recognition is divided into feature space recognition and input space recognition. The feature space recognition operates on signatures of images.) To convert images into templates:
  • Click “Search Dir” button to specify the search-directory.
  • Click menu item “BioFilter 2/Scan Images  - Directory Input” to convert images to records. You should see the ImageFinder scan through the images at this point.
    • “BioFilter 2/Scan Images  - File Input” Menu Item
    Use “BioFilter 2/Scan Images  - File Input” menu item to convert images into records. The BioFilter II is one of two filters operating on image segments in feature space images recognition (Image recognition is divided into feature space recognition and input space recognition. The feature space recognition operates on signatures of images.) To convert images into templates:
    “BioFilter 2\Train (match2.txt required)” Menu Item
    Use “BioFilter 2\Train (match2.txt required)” menu item to train the BioFilter II. Training uses the data collected in advance to teach the BioFilter II how to match. Training requires two files, d1.txt and match2.txt:
  • D1.txt is the record file, which contains many records. Each image is converted into a set of records. A record represents features of an image segment in a feature space.
  • Match2.txt is a list of matching pairs. This file will teach the ImageFinder who will match with whom.
    • “BioFilter 2/1:N Match (First vs. Rest)” Menu Item
    Use “BioFilter2/1:N Match (First vs. Rest)” menu item to make a 1:N Matching. 1:N Matching compares one key image with the images in a search-directory or search-file; the key image is specified in the “Key Segment” textbox or selected by the “Key Segment” button. 1:N Matching requires (1) the images in the search-directory or search-file being converted into templates in advance; and (2) the BioFilter II being trained. To make an 1:N Matching:
  • Click the “Key Segment” button, and select an image;
  • Click “BioFilter 2/1:N Match (First vs. Rest)”.
    • The result is in file, e1.txt, which will be opened at the end of computation.
    “BioFilter/N:N Match” Menu Item
    Use “BioFilter/N:N Match” menu item to make an N:N Matching. N: N Matching compares each image, specified in the search-directory or search-file, with every image in the search- directory or search-file. N:N Matching requires (1) the images in the search-directory or search-file being converted into templates in advance; and (2) the BioFilter II being trained. The result is in file, e1.txt, which will be opened at the end of computation.
     
    “BioFilter 2/BioFilter-2 Results” Menu Item
    Use “BioFilter 2/BioFilter-2 Results” menu item to open e1.txt, the file containing the last matching result.
    Parameters
     Please see the BioFilter section.


    23.9   Neural Filter II

    Neural Filter II Drop Down List

    Use the “Neural Filter II Drop Down List” to select a Neural Filter-II. The Neural Filter-II is the main matching filter operating on image segments for the Feature Space.
    “NeuralFilter 2\Train (match2.txt required)” Menu Item
    Use “NeuralFilter 2\Train (match2.txt required)” menu item to train the Neural Filter. Training uses the data collected in advance to teach the Neural Filter how to match. Training requires two files, d1.txt and match2.txt:
  • D1.txt is the record file, which contains many records. Each image is converted into a set of records. A record represents features of an image segment in a feature space.
  • Match2.txt is a list of matching pairs. This file will teach the ImageFinder who will match with whom.
    • “NeuralFilter 2/1:N Match (First vs. Rest)” Menu Item
    Use “NeuralFilter 2/1:N Match (First vs. Rest)” menu item to make a 1:N Matching. 1:N Matching compares one key image with the images in a search-directory or search-file; the key image is specified in the “Key Segment” textbox or selected by the “Key Segment” button. 1:N Matching requires (1) the images in the search-directory or search-file being converted into templates in advance; and (2) the NeuralFilter II being trained. To make an 1:N Matching:
  • Click “Key Segment” button, and select an image;
  • Click “NeuralFilter 2/1:N Match (First vs. Rest)”.
    • The result is in file, e1.txt, which will be opened at the end of computation.
    “NeuralFilter 2/ N:N Match” Menu Item
    Use “NeuralFilter 2/N:N Match” menu item to make an untrained N:N Matching. N:N Matching compares each image, specified in the search-directory or search-file, with every image in the search-directory or search-file. N:N Matching requires (1) the images in the search-directory or search-file being converted into templates in advance; and (2) the NeuralFilter II being trained. The result is in file, e1.txt, which will be opened at the end of computation.
    “NeuralFilter 2/NeuralFilter-2 Results” Menu Item
    Use “NeuralFilter 2/NeuralFilter-2 Results” menu item to open e1.txt, the file containing the last matching result.
    Parameters
     Please see the Neural Filter section.


    23.10   Neural Net II Filter

    Neural Net II Filter Drop Down List

    Use the “Neural Net II Filter Drop Down List” to set the Neural Net II Filter. The default filter is 100x100. All images are scaled down by an integer amount. For example, 640x480 will be scaled down 7 times to 91x68.
  • You need to understand it; so when you translate, or rotate an image, you will make sure no additional scaling factors are introduced.
  • The search speed crucially depends on this filter. For example, if the 50x50 filter is used, then the underlying neural net size is reduced by a factor of 4; and the neural computation speed will be increased by a factor of 16.
    • The filters available are:
  • 100x100
  • 90x90
  • 80x80
  • 70x70
  • 60x60
  • 50x50
    • Let the speed of 100x100 representation be a base, then the overall speed for:
  • 90x90 representation is 1 times faster;
  • 80x80 representation is 1.6 times faster;
  • 70x70 representation is 2.7 times faster;
  • 60x60 representation is 5 times faster; and
  • 50x50 representation is 10 times faster.
    • NeuralNet 2/1:N Match (First vs. Rest) Menu Item
    Use the "1:N Match (First vs. Rest)" menu item to retrieve images in the search-file. This command will train the software using the first image in the input file. After clicking the button, wait until a result file is opened.
    NeuralNet 2/NeuralNet-2 Results Menu Item
    Use “NeuralNet-2 Results” menu item to open the last matching result.
    Parameters
     Please see the Neural Net Filter section.


    23.11   Segment-Locator

    SegLocator/Train Menu Item

    Use “Train” menu item to train the Segment-Locator.
    SegLocator/1:N Match Menu Item
    Use “1:N Match” menu item to make a 1:N Match for the Segment-Locator using directory input.
    SegLocator/Batch Run Menu Item
    Use “Batch Run” menu item to make a batch run based on the batch code in the text area. The only valid Execution Code is 1027, therefore, you must use 1027, meaning 1:N Match using the Neural Net Filter and directory input.


    23.12   Other Objects In the Page

    Hints Button, Help/Hint Menu Item

    Use the "Hints" button or “Help/Help” menu item to see the basic procedure.
    Help Button, Help/Help Menu Item
    Use the “Help” button or “Help/Help” menu item to display the html version of this document.
    Clear Button, Help/Clear Window Menu Item
    Use the “Clear” button or “Help/Clear Window” menu item to clear the text area.
    Home Button, Help/Home Menu Item
     Go to http://attrasoft.com
    Image Display Area
    Use the “Image Area” to see the image(s). This area is 300 by 300 in size. If an image is chosen for specifying a segment, the image will be displayed in this area.
    Text Area
    Use the “Text Area” to see the current status of your execution, results, error messages, and suggestions.
    Help/About Menu Items
    Display email, web site, and version number.


    23.13   Examples

    Example/BioFilter

    Use this menu item to open batch codes for the following examples:

    N:N Match, Untrained, Label Template
    N:N Match, Trained, Label Template
    1:N  Match, Untrained, Label Template
    1:N  Match, Trained, Label Template
    1:N  Match, Trained, Label File
    N:N Match, Trained, Label File
    1:N  Match, Trained, Label Directory
    N:N Match, Trained, Label Directory

    Example/NeuralFilter
    Use this menu item to open batch codes for the following examples:

    N:N    Match, Label Template
    N:N-1 Match, Label Template
    1:N     Match, Label Template
    N:N    Match, Label File
    N:N-1 Match, Label File
    1:N     Match, Label File
    N:N    Match, Label Directory
    N:N-1 Match, Label Directory
    1:N     Match, Label Directory
    N:N    Match, Logo Template

    Example/Neural Net
    Use this menu item to open batch codes for the following examples:

    1:N Match, Logo, File
    N:N Match, Logo, File
    United Way - R
    Tabasco - R
    Mr. Potato - S
    Monopoly - S
    Compound - RS
    Stamp 1
    Stamp 2
    Long Search
    License Plates
    Fingerprint
    Fingerprint - 1
    Fingerprint - 2
    Fingerprint - 34
    Fingerprint - 7
    Fingerprint - 569

    Example/Segment Locator
    Use this menu item to open batch codes for the following examples:

     United Way
     Monopoly
     Mr. Potato
     AAA
    Ford
    Soup
     PointLocator: Feret 20
     PointLocator: Feret 100
     PointLocator: Results

    Example/BioFilter 2
    Use this menu item to open batch codes for the following examples:

    Label Match Setting
    Label Example 1 (1:N)
    Label Example 2 (N:N)
    Label Example 3 (1:N)

    Example/NeuralFilter 2
    Use this menu item to open batch codes for the following examples:

    Setting
    Example 1 (1:N)
    Example 2 (N:N)
    Example 3 (1:N)

    Example/Neural Net 2
    Use this menu item to open batch codes for the following examples:

    Match 1
    Match 2
    Match 3
    No Match 1
    No Match 2
    No Match 3

    Return