Getting started

Deepsight Face is currently supported in Linux and Windows Operating Systems on x64 platform.

It is available with or without GPU Acceleration. The free version comes without it.

Windows

Download and Install Deepsight Face

  1. Go to Deepsight Home and download the setup file.
  2. Run the setup file and install to a location that will NOT require admin privileges for writing. The default C:\Deepsight_Face is safe.

Install Dependencies

The setup package installs necessary dependencies. However, in case that didn’t happen, install these present in the redist folder.

  1. VC++ 2017 Runtime (vc_redist.x64.exe)
  2. Intel MKL BLAS Runtime (c_wproc*.exe)

Running Deepsight Face

  1. The setup should’ve created start menu links. Launch Deepsight Face from the start menu link. (or dsFace.exe from the installed folder)
  2. The application should start with a bunch of messages and finally say SERVER READY.
  3. At this point you can start the Demo app using your browser and pointing it to localhost:5000.

Linux

Download and Install Deepsight Face

  1. Go to Deepsight Home and download the compressed shell script file.
  2. Open a terminal and cd into a directory with non-root access.
  3. Copy the shell installer into this directory.

  4. Run the script using

    chmod +x Deepsight_Face-0.0.1-Linux.sh
./Deepsight_Face-0.0.1-Linux.sh

  5. Press the space bar to read the EULA and enter y to accept it

  6. Continue through the prompt until extraction is complete.

Install Dependencies

Deepsight on Linux requires the OpenBLAS library. It should be available in your distribution repository.

# On Ubuntu
sudo apt-get update
sudo apt-get install libopenblas-dev

Running Deepsight Face

  1. cd into the directory Deepsight_Face and use ./dsFace to launch the program
  2. The application should start with a bunch of messages and finally say SERVER READY.
  3. At this point you can start the Demo app using your browser and pointing it to localhost:5000.

Usage

  • The application accepts arguments as follows
$ ./dsFace -h
Deepsight Face is a Deep Learning powered face recognition SDK that runs locally as a http service
Usage:
  DeepSight Face [OPTION...]

  -v, --verbose     Print lots of messages
  -b, --benchmark   Run benchmark to evaluate speed
  -k, --key         Prompts license key
  -u, --usage       Print usage stats
  -f, --fastDetect  Uses faster hog face detector instead of slower mmod (with CUDA, mmod is faster)
  -h, --help        Prints help
  -p, --port arg    specify port at which to serve; default is 5000

Demo Application

The demo application provides a quick interface to try out all possible operations in Deepsight Face.

Launch

Fire up a browser now and point it to localhost:5000

Face Analysis

  • Switch to the first tab Face Analysis

Face Analysis Page

  • The interface is fairly intuitive. Select an image, check the attributes you would like to get, and press Analyze.

  • The returned json response is shown on the right. In case of error such as GPU memory full, you must relaunch the main application.

Face Comparison

  • Switch to the second tab Face Comparison

Face Comparison Page

  • The interface is fairly intuitive. Select two faces you would like to compare, and press Compare.

  • The returned json response mentions distance between the face embeddings in the higher dimensional space, threshold for equality (constant 0.6) and result of the comparison of distance with threshold.