This dataset contains all English words from the dataset available at MLCommons (or also available on huggingface). These audio files have been processed into Mel spectrograms for downstream usage in DCNNs or similar processes.
There's a total of 6624343 samples of Mel spectograms. There are a total of 38150 different words, the cls is the index of that word in alphabetical order. With every entry… See the full description on the dataset page:
https://huggingface.co/datasets/einstein8612/mlc-mlsw-melspects.
