Family Level Assessment of Screen Use in the Home-Television (FLASH-TV) Scripts

FLASH-TV installation, setup, configuration, and runtime scripts

Please refer to these publications for technical details:

• Vadathya et al. An Objective System for Quantitative Assessment of Television Viewing Among Children (Family Level Assessment of Screen Use in the Home-Television): System Development Study, JMIR Pediatrics and Parenting 5.1 (2022): e33569.

Installation Instructions

Please download the files from this Google Drive link and follow the below instructions

Prerequisites: Nvidia CUDA enabled GPU (8GB memory), Python3, PyTorch, MXNet

• Place the extracted .zip files from the download into the locations as specified in setup_scripts/file_setup.sh
• Run install_scripts/flash_install.sh to install all the components necessary for FLASH-TV (insightface, darknet face release)
• This creates a Python virtual environment at /home/$usrName/py38 from which we execute FLASH-TV algorithms

FLASH-TV Demo

• After the above installation steps are executed without any errors:
• Create a sample gallery of faces that you want to recognize. Take a look at the example gallery in examples/gallery_faces
• Please read the below instructions completely before executing FLASH-TV
• Run FLASH-TV v3.0 as indicated in runtime_scripts/run_flashtv_system.sh

Sample execution code For example, create a folder for saving data at /home/user/123_data. Create a sample gallery similar to the example and name it as 123_faces and put in the 123_data folder.

source /home/$usrName/py38/bin/activate
cd /home/$usrName/flash-tv-scripts/python_scripts
python test_vid_frames_batch_v7_2fps_frminp_newfv_rotate.py $famId $savePath save-image $usrName

• famId is a 3 digit unique ID for the specific data e.g. 123
• savePath specifies the location where data is saved e.g. /home/user/123_data
• save-image indicates whether to save image outputs, to not save specify no-save-image
• usrName indicates the user name of the user account

Details for data collection - FLASH-TV Installation on NVIDIA Jetson AGX Orin Devkits

Please read the setup instructions and Home Assistant usage files above for details

Make sure you have downloaded or copied the necessary software (insight_face, darknet_face_release etc.) and placed it in the right locations(*placement only, when specific external hard disk is connected)

• full_initial_configuration.sh

  Bash script that automatically runs or sets up other bash scripts for procedures marked with an asterisk(*), needs to be run once on every new Jetson

• install_scripts

  Bash scripts for installing ML libaries and packages necessary for running FLASH-TV and Home Assistant on the Jetsons*

FLASH-TV In-Home Data Collection

Please see the procedure and slides for detailed instructions. The procedure is also hosted as a pdf in this repository and updated periodically.

Bash scripts for running initial FLASH-TV protocols for setup, including building the gallery, creating faces, and running TV gaze estimation with images saved

• setup_scripts

  Bash scripts for the initial setup of the Jetsons in order to be able to run the other scripts and services and properly save data*

• services

  Services for systemd and bash scripts that will run FLASH-TV and Home Assistant on boot and restart FLASH-TV periodically(*setup only)

  Enable, start, and check the status of all relevant services quickly with test_services.sh if necessary

  Disable and stop all relevant services quickly with stop_services.sh if necessary

• participant_change.sh

  Bash script that automatically runs other bash scripts for changing the setup (IDs, services) when changing participants

• video_capture_scripts

  Bash scripts used on a laptop for video data collection for human ground truth labeling, deleted on Jetsons(*deletion only)

• python_scripts

  Python scripts for running FLASH-TV detection, recognition, and target child's gaze estimation