Barn preprocessing & upload
DialloLab at UQAM

×
Barn preprocessing & upload - DialloLab at UQAM

Project description:

A more advanced version of the McGill barn video uploads app. This desktop app form preprocesses the video data on-site (a barn) before uploading video files to the WELL-E file storage server. On-site preprocessing consists of multiple operations such as anonymization of barn staff, video trimming, video encryption, etc. to reduce the size of files transferred and to keep security at a maximum. This app uses Globus endpoints to transfer data efficiently as well as private packages to run preprocessing operations locally.

Tools & packages used:

  • Python
  • Flet
  • SQLite3
  • PyTorch
  • YOLOv7
  • OpenCV
  • Pandas
  • Seaborn
  • Globus endpoints

Responsibilities:

  • Design & build the GUI
  • Database design and implementation
  • Backend usage of code from other private packages
  • Automate preprocessing operations depending on the information filled in the form
  • Automate metadata gathering and remote database update depending on the information filled in the form

Other contributors to the project:

  • François Gonothi Toure

Code repo
McGill barn video uploads
DialloLab at UQAM

×
McGill barn video uploads - DialloLab at UQAM

Project description:

A simple desktop app form for uploading whole folders to the WELL-E file storage server. Under the hood, it leverages the rsync linux command from a Windows Subsystem Linux (WSL) instance to upload data over SSH. Using a desktop GUI eases the upload of data by biologists gathering data in the barn.

Tools & packages used:

  • Windows Subsystem Linux (WSL)
  • rsync
  • Python
  • PySimpleGUI
  • auto-py-to-exe

Responsibilities:

  • Design & build the GUI
  • Backend operations for leveraging rsync
  • Automated folder creation on server depending on the information filled in the form

Code repo
LLaMARJO app
DialloLab at UQAM

×
LLaMARJO app - DialloLab at UQAM

Project description:

Livestock labeling Learning for Manual Analysis & Recognition as a Journal of Observations
LLMARJO is a software that enables user-centric learning in the processes of behavior annotation.
This software is based on knowledge acquired manually in past experiments to provide automations in the process of video creation for training, practicing and certification as well as management of the users training.
LLaMARJO was created to ease the training process of behavior encoding in videos and remove pressure from the domain experts in the WELL-E team by automating certain tasks.

Tools & packages used:

  • Google AppSheet
  • Python
  • Pandas
  • FFmpeg
  • Google Drive API
  • Youtube API
  • Speech T5 by Microsoft from Hugging Face Hub

Responsibilities:

  • Global design
  • Data preprocessing
  • Database building
  • Data storage in Google Drive

Other contributors to the project:

  • Momar Aly Dom Fall

Code repo