Egocentric Video Segmentation

My (Max Wilson) Final Year Honours Project - Software used to segment a first-person video into Transitions and Visits

Requirements

• Install Python 3.8.2
• Install pip (or your own package manager)
  • Please run pip install -r requirements.txt in order to install the necessary requirements, or again through your own package manager.
• OpenCV 4.2.0

OpticalFlow

How to use

Using Powershell/Command Prompt, run:

python of.py "./Video/Path/Here.mp4"

Which will then output a file called opticalflow_results.csv to be used in the Classifier script.

Screenshots

Note: Time stated above is for a 40 minute video, your time may vary depending on computer specifications!

Classifier

How to use

Using Powershell/Command Prompt, run:

python analysis.py "TrainDataset" "GroundTruth_Train" "TestDataset" "OutputFile"

Where:

• TrainDataset: The optical flow file for the training data used to train the model (e.g. opticalflow.csv)
• GroundTruth_Train: The GroundTruth Data for the supplied optical flow file (.EAF annotated file e.g. annotate.eaf)
• TestDataset: The optical flow file for the video you wish to be annotated (test_opticalflow.csv)
• OutputFile: The name and/or location of the file you wish to be outputted (a .EAF file e.g. output.eaf)

Note: Please ensure you have the template.eaf file downloaded and in the same directory when running the script!

Screenshots

Using the Data

Once both of these scripts have run successfully, you will be left with a file YourNameHere.eaf. Open this with the program ELAN to view the completed annotated video. You will need to link the location of the video within the application in order to view it correctly.

To do this:

• Open YourFile.eaf in ELAN
• Press Edit then Linked Files... (or use the key combination CTRL+ALT+L)
• If there are any files listed under the Linked Media Files tab, remove them and click Apply
• Press Add and select your video file
• Apply then Close