MM Data Set Downloader

A Python library for downloading various multimodal datasets used in machine learning research. This tool simplifies the process of obtaining datasets for multimodal analysis and experimentation.

Installation

You can install this package using pip:

pip install git+https://github.com/jmg049/DataSets.git

Usage

After installation, you can use the mm_dataset command-line tool:

mm_dataset --dataset DATASET_NAME [--download_dir DOWNLOAD_DIR] [--unzip] [--del_zip]

Options

--dataset: Required. Dataset to download. Choices are: ["mmimdb" "avmnist" "kinetics-sounds" "mosi" "mosei" "msp-improv" "iemocap" "genspeech" "tcd_voip" "audioset_urban_rural"]

--download_dir: Directory to save the dataset (default: "data")

--unzip: Flag to unzip the downloaded file (default: False)

--del_zip: Flag to delete the zip file after unzipping (default: False)

Available Datasets

The following datasets are supported:

• MMIMDb: Multimodal IMDb dataset
• AVMNIST: Audio-Visual MNIST dataset
• Kinetics-Sounds: Audio-visual dataset for action recognition
• MOSI: Multimodal Opinion Sentiment Intensity dataset
• MOSEI: Multimodal Opinion Sentiment and Emotion Intensity dataset
• MSP-IMPROV: Multimodal database of improvisational emotional dyadic interactions
• IEMOCAP: Interactive Emotional Dyadic Motion Capture database
• GenSpeech: Generated Speech dataset
• TCD-VOIP: Voice Over IP Quality dataset
• AudioSet Urban-Rural: Urban and rural sounds from AudioSet (wav only)

Example

To download the MMIMDb dataset and save it in the "data" directory:

mm_dataset --dataset mmimdb --download_dir data --unzip --del_zip

This command will download the MMIMDb dataset, unzip it in the "data" directory, and delete the original zip file.

License

This project is licensed under the MIT License.