examples

DeepLib Examples

This directory contains example scripts demonstrating how to use DeepLib for various deep learning tasks.

Prerequisites

Before running any example, make sure you have installed DeepLib and its dependencies:

# Install PyTorch and torchvision first (required)
pip install torch torchvision

# Install DeepLib
pip install deeplib  # for latest release
# or
pip install -e .    # for development installation

Available Examples

Semantic Segmentation (train_segmentation.py)

A complete example showing how to train a UNet model for image segmentation tasks.

Dataset Structure

Your dataset should be organized as follows:

data_root/
├── images/
│   ├── train/
│   └── val/
├── masks/
│   ├── train/
│   └── val/

Basic Usage

python train_segmentation.py \
    --data_root ./data/segmentation \
    --num_classes 3

Advanced Options

Option	Description	Default
--data_root	Root directory containing the dataset	Required
--images_dir	Directory name containing images	"images"
--masks_dir	Directory name containing masks	"masks"
--num_classes	Number of segmentation classes	Required
--num_epochs	Number of training epochs	50
--batch_size	Batch size for training	64
--learning_rate	Learning rate	1e-4
--input_size	Input image size	192
--ignore_index	Index to ignore in loss calculation	255
--dropout_p	Dropout probability	0.1

Loss Functions

Choose from multiple loss functions using the --loss argument:

• ce: Cross Entropy Loss (default)
• dice: Dice Loss
• wce: Weighted Cross Entropy Loss
• jaccard: IoU Loss
• focal: Focal Loss

Example with Dice loss:

python train_segmentation.py \
    --data_root ./data/segmentation \
    --num_classes 3 \
    --loss dice

Experiment Tracking

Track your experiments using different loggers with the --logger argument:

• tensorboard: TensorBoard logging (default)
• mlflow: MLflow logging
• wandb: Weights & Biases logging
• none: No logging

Example with W&B logging:

python train_segmentation.py \
    --data_root ./data/segmentation \
    --num_classes 3 \
    --logger wandb