Term-Annotator

General

Term-Annotator is an application for media (video/images/gifs) annotation.

This application is installed on a laptop and is working locally. It is made for images/video/gifs manual classification in order to train machine learning on the database.

Basically, term annotator is only moving files from one directory to another in a local computer.

Main purpose

Term-annotator provides productivty tools for the annotator :

• Visualization of images/gifs/videos
• Keyboard shortcuts for each category
• Ability to split videos into sub-videos
• Ability to organize categories using Questions/answers tree
• Thumbnails display for each categories
• Display of the number of items in each category in real-time

Usage

1. Load application with this link and unzip into term-annotator folder

2. Load mediafiles to annotate and unzip into new folder

3. Run term-annotator.exe file from term-annotator folder

4. Import Configuration file:

   4.1. Click on button Import Configuration File

   

   4.2. in a new window CHOOSE CONFIGURATION FILE choose your config file.

   

   4.3. in a new window CHOOSE SOURCE DIRECTORY choose folder with files to annotate

   

5. First image is displayed to start annotation.

6. when annotation of first image is done, annotated image will be automatically saved in newly created annotated folder and removed from original folder where all the images are saved.

7. In annotated folder, image is being stored in folder with name-patterns like : <team-number>-<player-number> and notVisible-notVisible-color along with notPlayer and other folder.

8. Generate Annotation results

   8.1 click on button Generate Result

   8.2 save result as json file. See naming pattern here

Detailed features

Gifs Thumbnails

• thumbnail is created once first player of category is annotated
• thumbnail cam be updated with a click on button NextImage.

Configuration

Configuration parameters for the term-annotator are strored in a json file that should be loaded once app is started.

Configuration file should be shared with the annotator together with a data set before the annotation

EXAMPLE

{
    "sort": true,                               
    "source" : {
        "dir": "my_directory",
        "recursive": true,
        "mediaType": "video",
        "subDir":"toannotate"
        "videoSplit": {
            "startEndRegexp" : "\\w*-([0-9\\.]*)-([0-9\\.]*).mp4",
            "splitKey": "!",
            "minDuration": 2
        },
        "renaming": {
            "from": "/",
            "to": "_xxx_"
        }
    },
    "destination": {
            "destSubDir":"annotated"
        },
    "classes": {
        "class-a": {
            "key" : "a",
            "dir": "myAnnotation/class-a"
        },
        "class-c": {
            "key" : "c",
            "dir": "myAnnotation/class-c-<%=isFoo%>-<%=isBar%>",
            "params" : [{
                    "label": "Is it foo ?",
                    "name": "isFoo",
                    "choices": [{
                            "key": "Y",
                            "label": "yes",
                            "value": "fooTrue"
                        },{
                            "key": "N",
                            "label": "no",
                            "value": "fooFalse"
                    }]
                }
                }]
        }
    }
}

Import/Export Results

Results format

{
  "epoch": "2019-10-25T11:13:02.723Z",
  "folders": {
    "dir-path" : {
            "thumbnail": null,
            "filenames": [
              "a"
            ]          
     },
  "termAnnotatorConfig": {...}   
     ....
}

epoch - an import timestamp

dir-path - a path to dir for a category. Each annotated category has separate directory

filenames - a list of files inside dir

thumbnail - a thumbnail stored inside dir

termAnnotatorConfig - a configuration for the correction workflow

Example

{
  "epoch": "2019-10-25T11:13:02.723Z",
  "folders": {
    "../annotated/1-5" : {
            "thumbnail": null,
            "filenames": [
              "a",
              "b",
              "c",
              "w"
            ]          
     },
     ...
    "../annotated/notClean":
      {
        "filenames": [
          "q",
          "p",
          "r",
          "s"
        ],
        "thumbnail": s.png
      }  
   },
   "termAnnotatorConfig": {
    "sort": true,
    "mediaType": "gif",
    "source": {
        "subDir": "toannotate",
        "recursive": false
    },
    "destination": {
        "destSubDir": "annotated"
    },
    "classes": {
        "player01": {
            "key": "1",
            "dir": "1-1_TANx1-1_SMI"
        }
    }
}

Export Result

1. click on button Generate Result
2. save result as json file. See naming pattern here

Import Result

1. Go to app's Menu -> Advanced -> Import Annotation File
2. Choose a Result file to be exported in new window
3. Choose a Source Directory in new window (<name-of-dataset/toannotate>)
4. Choose an Destination directory in new window (<name-of-dataset/annotated>)

Packaging

Local

To package repo into application run:

- for Windows : `npm run package-windows`
- for Ubuntu: `npm run package-linux`

Packaged app is stored in:

- for Windows : `./term-annotator-app-linux.zip`
- for Ubuntu: `./term-annotator-app-windows.zip`

Release into GitHub assets

Push new changes into master to release new version of app with Semantic-Release

Download from GitHub

Download app from Releases' assets: https://github.com/teamklap/term-annotator/releases

• for Windows : download term-annotator-app-windows asset
• for Ubuntu: download term-annotator-app-linux asset

Start app

1. unzip packaged app:

   • for Windows : unzip with zip manager WinRAR
   • for Ubuntu: unzip term-annotator-app-linux.zip -d .

2. run app:

   • for Windows: term-annotator-win32-x64/term-annotator.exe
   • for Ubuntu: term-annotator-linux-x64/term-annotator