Skip to content

AF Tool

Introduction

The IDLive Doc Anonymized Feedback Tool (hereafter AF tool) is intended for collecting business intelligence data for customers' document photo datasets. The tool is designed to be run by customers on customers' premises without a need to send images over external network. The AF Tool is shipped as a standalone Docker image. Hardware and software requirements for it are the same as for Docker distribution of the IDLive Doc product.

Data privacy

  1. The AF Tool does not send any data to ID R&D by itself and does not require internet connection.
  2. The output of the AF Tool processing is a plain text CSV file which is intended to be downloaded and inspected from a privacy perspective. ID R&D encourages users to review the content and redact any information they do not wish to transmit.
  3. The AF Tool does not include any personal identifiable information (PII) in its report (e.g., filenames). None of data present in the AF Tool report can be used for reconstructing source images.

Here is the content and a description of each algorithm in the results:

  • Errors - Indicates any errors produced by the Docker while processing submitted image
  • Image Width and Image Height - Dimensions of the submitted image
  • Image Mode - RGB or other pixel arrangement formats
  • Image Format - JPEG or other image formats
  • Image Orientation - Landscape or Portrait image capture
  • Image EXIF Compression - if any compression is indicated in any included EXIF data
  • Image EXIF Orientation - if orientation tag is specified in EXIF data
  • Image Compression Quality - image compression quality estimated using quantization tables
  • Image Histogram - Displays the standard Image Histogram information as described here. This information cannot be used to reconstruct the original image, and is used by ID R&D to assist with confirming the image is within tolerances for high quality results
  • Bounding boxes - Locations of Corners of the document or portrait within the image
  • Black and White Document Validator - Checks whether the image is all in black and white without color
  • Quality Check - Detection - Lists the detection quality warnings found
  • Quality Check - Image - Lists the image quality warnings found
  • Model Results - These data fields return raw results from individual machine learning models. These values are combined in various ensembles which result in the pipelines delivered through our core product. Transmitting these values helps ID R&D determine whether some model sub-components cover domains which are not covered by the pipelines delivered through our core product. The values in the generated CSV are truncated for additional data security.

Basic usage

Note

To provide feedback, please send us the resulting CSV from one or more runs of the tool. If you have spoof and genuine images, please send a separate CSV for each. If you are concerned about testing across document types, please send a CSV of images from each document type.

This step-by-step guide assumes that you have already install Docker engine and received the AF Tool Docker image from ID R&D. So that, the docker image ls command output should look like this:

$ docker images
REPOSITORY                    TAG        IMAGE ID       CREATED        SIZE
idlivedoc-anon-fedback-tool   1.3.0      7fbaf67ed086   20 hours ago   2.5GB

Below guide will walk you through the end-to-end process of obtaining the AF Tool report:

  1. Run the AF Tool Docker image exposing the 8000 port of the container and wait until the initialization is completed (Initialized! row in the log output):

    $ docker run -it -p 8000:8000 idlivedoc-anon-fedback-tool:1.3.0
Started License Manager
License /app/trial-license-2022-11-22.adminmode.v2c installed
INFO:logger:Initializing...
INFO:logger:Initialized!
INFO:     Started server process [28]
INFO:uvicorn.error:Started server process [28]
INFO:     Waiting for application startup.
INFO:uvicorn.error:Waiting for application startup.
INFO:     Application startup complete.
INFO:uvicorn.error:Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
INFO:uvicorn.error:Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

  2. Go to the http://127.0.0.1:8000 URL in a web browser. You should see the AF Tool UI.

  3. Using the provided file upload form, add image files or folders for analysis. You can do it either by drag-and-drop from the file explorer or using manual file or folder selection provided by “Choose image files / folder“ buttons.

  4. After submitting image files for processing you should see the information on processing progress. Once processing is done, the “Download report“ button become available.

    If you want to submit more images for processing, just add more files the same way.

    If you are done with submitting images and you want to obtain the AF Tool report in CSV file, press the “Download report“ button. You will be offered to download the CSV file. This action also resets the processing session, so the processing results for newly submitting files will be held in another report.

    If you want to drop all results without downloading the report, press the “Reset session“ button.

Advanced usage

In addition to the web UI, the AF Tool also provides REST API available in the same distribution. After running the Docker container you can study the API entries by visiting the Swagger UI API reference at http://127.0.0.1:8000/api/docs.

Below example describes basic usage of the API using curl utility:

$ curl -X GET http://127.0.0.1:8000/api/sessions/create
"85eef64c-2777-49a7-ac4e-8eb6161599a8"
$ curl -X POST -F file=@image.png http://127.0.0.1:8000/api/sessions/85eef64c-2777-49a7-ac4e-8eb6161599a8/process_file
"OK"
$ curl -X GET http://127.0.0.1:8000/api/sessions/85eef64c-2777-49a7-ac4e-8eb6161599a8/finalize > report.csv
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  1455    0  1455    0     0   481k      0 --:--:-- --:--:-- --:--:--  710k

Troubleshooting

If you suspect that UI is stuck and throwing errors like “Session not found“, try deleting the web cookies for the AF Tool UI URL in your web browser.