Our project is MODLBOX, an application to speed up the Machine learning/deep learning development lifecycle. It aims to provide an intuitive user interface and is helpful for people with limited coding experience.
MODLBOX integrates various machine learning libraries, thus enhancing flexibility and performance in model training. Users can effortlessly select different components of the DL pipeline, including databases, and models, to create end-to-end models.
Requirements:
- Creating/Arranging the folder structure as per the model.
- Ability to generate graphs.
- Integrating terminal in the application.
- Option to select between classification (labeling an image) and object detection (identifying an object within an image).
- Dataset Selector(Standard dataset), different dataset selector for classification and detection.
- Option to select different models. I.e Alexnet, Yolo V8, Yolo V5 etc.
- Option to select GPU. In order to review configurations effectiveness with and without GPU constraints
- Option to set parameters, i.e number of Epochs to train, Batch size.
Sotwares Used:
- PySide6
- Tensorboard
- Ultralytics
- Pyinstaller
- Pytorch
Programming Language:
Python (Frontend & Backend)