Changes between Version 5 and Version 6 of Event/2012/GCI/QA

11/02/12 22:01:41 (11 years ago)
Mark Prutsalis



  • Event/2012/GCI/QA

    v5 v6  
    4343== Sahana Vesuvius Tasks ==
     45=== Draw or Evaluate Annotations of Faces using "!ImageStat" Tool ===
     46Research is going on to improve face-detection and face-matching algorithms, for eventual incorporation into Vesuvius/PL and other projects.  To assist with algorithm training/testing and "ground truthing", images of faces are being annotated with rectangles.  A web-site tool, "!ImageStat", has been built to allow this.  Annotators are needed to use this tool on public images, e.g., of celebrities.
     48The specific task (broken into finite 2-3 hours chunks) may involve:
     49* Drawing a frontal "face" rectangle, a "profile" rectangle, and/or representative "skin" rectangles.
     50* Comparing such rectangles drawn by hand to rectangles marked by computer, and judging their similarity.
     52=== Annotate or Evaluation Extraction of Person Names, Place Names and Health Status of Missing and Found Persons ===
     54Research is underway to apply computer methods to extracting key features related to missing/found persons from English sentences found in free text.  The sentences may be artificial or from a collection.  Tasks may involve:
     56* Identifying certain parts of speech within a sentence, as “ground truth” tagging.
     57* Identifying person names, place names, and health status within a sentence, as “ground truth” tagging.
     58* Evaluating the accuracy of parts-of-speech tagging done by computer parsing of sentences.
     60There will likely not be task-specific software for this activity, rather the use of standard office tools like spreadsheets.  It will be helpful to be a native or very fluent English speaker with some exposure to formal grammar rules.
    4664[wiki:Event/2012/GCI GCI]