To augment your dataset of mel-spectrograms in order to increase its size for deep learning model implementation you can try imgaug Tensorflow, Datacamp

PyTorch has multiple built-in methods for data augmentation, which are especially handy if you are implementing your DNN in PyTorch.

See, e.g.,

https://www.kaggle.com/code/mohamedmustafa/7-data-augmentation-on-images-using-pytorch

I suggest to read, https://www.datacamp.com/tutorial/complete-guide-data-augmentation

Hey Jyoti Mishra , for the mel-Spectrograms use the TensorFlow's Image Data Generator to perform augmentation.

ImageDataGenerator- https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator

If you are comfortable with reading docs and Python's layer mapping syntax( using .map() to map Augmentation layers on batches of data), I would recommend trying Keras CV.

Keras CV-

https://keras.io/guides/keras_cv/cut_mix_mix_up_and_rand_augment/

Certainly! There are several libraries in Python that you can use for data augmentation of RGB images. One popular library is OpenCV, which provides a wide range of image processing functions, including resizing, rotation, flipping, and adding noise. Another powerful library is imgaug, which offers a comprehensive set of augmentation techniques such as cropping, scaling, rotation, and color transformations. Albumentations is another popular choice that supports a diverse range of augmentation techniques and is known for its fast execution speed. Additionally, keras.preprocessing.image module provides convenient functions for image augmentation, including rotation, shift, and zoom. By utilizing these libraries in your Jupyter Notebook, you can simplify the process of augmenting your mel-spectrogram dataset, enabling you to effectively increase its size for deep learning model implementation.