deepencoding / face-clipper

Read the mp4 video file and write JPEG encoded frames. Read each JPEG frame one by one and detecting faces from the image and crop it.

1. A simple gstreamer pipeline and run using gst-launch command line tool to read the mp4 video file, scale down the input to 640x640 resolution and write JPEG encoded frames.

1:

1. cd into 1_stream_to_frames directory.

2. Run the shell script and pass two command line arguments:
   argv[1]: Path to the mp4 file (I used ./sample_video/sample_1.mp4)
   argv[2]: Name of the output folder (I used out_1 for sample_1.mp4 and out_1_2 for sample_3.mp4)

   So it looks something like this:
   $ ./stream_to_frames.sh ./sample_video/sample_1.mp4 out_1

   The output will be in the same directory with the name you gave when running the command.

2: A C++ application to read each JPEG frame one by one. Uses OpenCV’s pre-trained XML classifier for detecting faces from the image and crop it. Creates a frame wise folder structure of cropped faces.

2:

1. cd into 2_CaptureFaces directory.

2. Run the command: $ cmake . To build the VS Solution.

   note: If you want to change the directory from which the C++ application takes input of the frames from, let's say you want to change the folder from 1_stream_to_frames/out_1_2 to 1_stream_to_frames/out_1"":

   1. Open CMakeLists.txt in a file editor.

   2. Locate the set_target_properties(CaptureFaces PROPERTIES VS_DEBUGGER_COMMAND_ARGUMENTS "../1_stream_to_frames/out_1_2/") command and change the command line argument:
      set_target_properties(CaptureFaces PROPERTIES VS_DEBUGGER_COMMAND_ARGUMENTS "//Enter changed directory here.//")

      Let's say if you want to take input from out_1 in the 1_stream_to_frames directory, then it will look something like this: set_target_properties(CaptureFaces PROPERTIES VS_DEBUGGER_COMMAND_ARGUMENTS "../1_stream_to_frames/out_1/")

      save and exit and start from step 1.

3. Open the 'CaptureFaces.sln' file in Visual Studio and there you have the C++ Application project.