This document is under construction. It will describe the details about integrating already generated metadata with Clowder.
Background
At the time of writing this document, there are about 171,000 images from the Library of Congress Farm Security Administration/Office of War Information Photograph Collection that were processed for extracting various features like faces, eyes, facial profile, closeups, printed text, presence of Stryker hole, presence of border, mean and standard deviation of grayscale values, subject details, photographer details, and category details. These were computed on XSEDE Comet by using stripped down version of Clowder Extractors or in certain cases, by creating new standalone programs. Integrating this information with Clowder is important to use its features like RESTful API, authentication and authorization, available visualizations, etc.
Database Table Descriptions and Converted Sample JSON Documents
The following set of tables contain description about the database tables that contain extracted metadata. Following each database table description is an example JSON document that will be generated by the Extractor Integration Script for that particular table.
CategoryInfo
| Sl. No. | Database Column Name | Field Description | Remarks |
|---|---|---|---|
| 1 | id | LOC Index | String; |
| 2 | category | LOC Category number (other_number field in the image JSON document) | String; |
{
"id": "fsa1997018591",
"category": "F 665"
}
CreatorInfo
NOTE: some creators are empty strings, so it might need some refinement.
| Sl. No. | Database Column Name | Field Description | Remarks |
|---|---|---|---|
| 1 | id | LOC Index | String; |
| 2 | name | Creator name in the format: <last name>, <first name>, <birth year> - <death year>. If the creator name is blank the value is NULL. | String; |
| 3 | year_mon | Year and month (abbreviated in certain cases) in which the photograph was taken in the format: <year> - <month | month1 - month2 | season > | String; Some year - month values are like '[between 1940 and 1946]'; The format mentioned in the left cell may not be strictly followed. Need to look into this in detail when doing the transformation. |
FacesInfo
| Sl. No. | Database Column Name | Field Description | Remarks |
|---|---|---|---|
| 1 | id | LOC Index | String; |
| 2 | imght | Image height | Float; |
| 3 | imgwid | Image width | Float; |
| 4 | dumb1 | The letter F, it's only there to help browse raw data | String; |
| 5 | num_faces | Number of faces found | Integer; |
| 6 | face_segs | Bounding box location of faces | String; this is a text string that has the ith face, x, y, width, height of face segment. Each face segment is separated by a semicolon. |
| 7 | dumb2 | The letter P, it's only there to help browse raw data | String; |
| 8 | num_profiles | Number of profiles found | Integer; |
| 9 | prof_segs | Bounding box location of profiles | String; |
| 10 | dumb3 | The letter Y, it's only there to help browse raw data | String; |
| 11 | num_eyes | Number of eyes found | Integer; |
| 12 | eye_segs | Bounding box location of eyes | String; |
| 13 | dumb4 | The letter C , it's only there to help browse raw data | String; |
| 14 | num_fullcls | Number of face full closeups | Integer; 'FULL' is relative to image size |
| 15 | num_midcls | Number of face mid closeups | Integer; 'MID' is relative to image size |
| 16 | num_fullprof | Number of profile full closeups | Integer; 'FULL' is relative to image size |
| 17 | num_midprof | Number of profile mid closeups | Integer; 'MID' is relative to image size |
ImageProperties
| Sl. No. | Database Column Name | Field Description | Remarks |
|---|---|---|---|
| 1 | id | LOC Index | String; |
| 2 | hole | Presence of Stryker hole | Boolean; |
| 3 | border | Presence of border | Boolean; |
| 4 | meangray | Mean of grayscale values (not including hole and border) | Float; |
| 5 | stdgray | Standard deviation of grayscale values (not including hole and border) | Float; |
ImageFilesList
| Sl. No. | Database Column Name | Field Description | Remarks |
|---|---|---|---|
| 1 | fileid | File ID (Serial number) | Integer; |
| 2 | id | LOC Index | String; |
| 3 | cometfn | Filename in Comet | String; |
| 4 | locurl | URL of the photograph in LOC website | String; |
OCRInfo
| Sl. No. | Database Column Name | Field Description | Remarks |
|---|---|---|---|
| 1 | id | LOC Index | String; |
| 2 | ocr_pred | Overall prediction of whether or not text is present in image. 'nop' means OCR found nothing. Where if any one box predicted text then the final prediction is set to T. | String; What is the difference between nop and F? |
| 3 | scores | Prediction scores. A string that consists of sets of 3 numbers (separated by semicolon) where, for each OCR text box found, a 0/1 classification value indicating no-text/text predicted, 2 floats indicating classification score for no-text/text | String; |
| 4 | box_sum | A count of number of 1's found across text box score sets | Integer; |
| 5 | box_cnt | Number of text boxes. Note that box_sum / box_cnt is another possible score instead of the T/F above. | Integer; What is the difference between box_sum and box_cnt? |
| 6 | box_txt | Set of strings separated by semicolon. One string for each text box found in OCR process | String; |
| 7 | box_locs | A string that consists of sets of 4 numbers (separated by semicolon) one set for each text box, where the numbers are upper left x coordinate, upper left y coordinate, box width, box height. | String; |
SubjectInfo
| Sl. No. | Database Column Name | Field Description | Remarks |
|---|---|---|---|
| 1 | id | LOC Index | String; |
| 2 | subject | Subject information | String; |