Multi-Label Image Classification using Pytorch

MultiLabelImageClassificationPytorch is a robust and flexible library designed to simplify the process of multilabel image classification with a dataset of images. This library provides a suite of scripts and modules to load various models, fine-tune hyperparameters, and even train multiple models sequentially with ease. The project boasts in-depth visualization support through TensorBoard, enabling users to compare performance across models and optimize their specific use cases efficiently.

Table of Contents

• Installation
• Usage
• Understanding Results
• Project Structure
• Contributing
• License

Installation

To set up the environment to run the code, follow these steps:

Prerequisites

• Anaconda or Miniconda

Environment Setup

1. Clone the repository to your local machine:

   git clone https://github.com/SkierProjects/MultiLabelImageClassificationPytorch.git
   cd MultiLabelImageClassificationPytorch

2. Create and activate the conda environment from the environment.yml file:

   conda env create -f environment.yml
   conda activate multilabelimage_model_env

Usage

Dataset Preparation

Place your dataset.csv file in the Dataset directory. The CSV file should have the following format:

filepath,classname0,classname1,...
/path/to/image1.jpg,0,1,...
/path/to/image2.jpg,1,0,...

Run analyzeData.py to get insights on the class balance in the dataset:

python Dataset/analyzeData.py

Run computemean.py to calculate the mean and standard deviation of your dataset

python src/computemean.py

Take the outputs for the mean and standard deviation and place them inside src/config.py for dataset_normalization_mean and dataset_normalization_std. These will be used to normalize the images used for training.

Training a Model

To train a model, use the train.py script:

python src/train.py

For training multiple models, use the train_many_models.py script. Modify train_many_models.json to include the models you want to train, which overrides values in config.py:

python src/train_many_models.py

Evaluating a Model

To evaluate the performance of a model, you can use the test.py script:

python src/test.py

Evaluating a Model

To run a model and get results at runtime, you can use the inference.py script:

python src/inference.py

TensorBoard

To view TensorBoard logs, run:

tensorboard --logdir=tensorboard_logs

Understanding Results

The results of the model training and evaluation will be stored in the following directories:

• logs: Contains log files with detailed information about the training process.
• outputs: Contains saved models in .pth format.
• tensorboard_logs: Contains TensorBoard logs for visualizing training progress and metrics.

Project Structure

• Dataset/: Contains the dataset and scripts for analyzing the dataset.
• logs/: Where training logs are stored.
• outputs/: Where trained model weights and checkpoints are saved.
• tensorboard_logs/: Where TensorBoard logs are output.
• src/: Contains all source code for the project.
• environment.yml: Conda environment file with all required dependencies.

Contributing

To contribute, please submit a pull request to the repository. Your contributions will be reviewed and considered for merging.

License

This project is licensed under the Apache License 2.0. See the LICENSE file for more details.