search

MiteshPuthran / Speech-Emotion-Analyzer

The neural network model is capable of detecting five different male/female emotions from audio speeches. (Deep Learning, NLP, Python)
MIT License
1.31k stars 438 forks source link

Dataset question #14

Closed hackiey closed 5 years ago

hackiey commented 5 years ago

Hi, thanks for the work, I have a question about the samples number. After I filter out all the data (ravdess and savee), I just have 1200 samples, 960 train data, and the final acc is around 0.5. I found your train samples is 1378 (X_train.shape is 1378*216), I wonder what I did wrong.

MiteshPuthran commented 5 years ago

Hi @hackiey, it's just the number of files that I had in the dataset. They must have changed the number of audio files in their website.

hackiey commented 5 years ago

Hi, I calculated the number of the dataset, in the RAVDESS, there are 96 samples of neutral, and 192 samples in each of the other categories, a total of 1500 samples, I got 192 x 5 = 960 samples from this dataset. in the SAVEE, there are 60 samples per category, a total of 480 samples, I got 60 x 4=240 samples from this dataset. Because there are about 2,000 samples in total, the number of our data sets should be the same.

MiteshPuthran commented 5 years ago

I am having around 2880 files in my dataset. Including data from both the sources

hackiey commented 5 years ago

wow, maybe they change the dataset, how can I get this data? or could you upload this data?

Arunpoochelvan commented 5 years ago

Can you upload the dataset?

Glastond commented 5 years ago

Yeah please... because it's creating a lot of confusion for me... Uploading the database will help me a lot! By the way great work!

Arunpoochelvan commented 5 years ago

Yeah please... because it's creating a lot of confusion for me... Uploading the database will help me a lot! By the way great work!

I got it brother, Ref this link you got all the dataset https://github.com/preranas20/Emotion-Detection-in-Speech/tree/master/Emotion%20Detection%20in%20Speech ROCK AND ROLL

Arunpoochelvan commented 5 years ago

wow, maybe they change the dataset, how can I get this data? or could you upload this data?

I got it brother, Ref this link you got all the dataset https://github.com/preranas20/Emotion-Detection-in-Speech/tree/master/Emotion%20Detection%20in%20Speech ROCK AND ROLL

drbln commented 5 years ago

Hi, I calculated the number of the dataset, in the RAVDESS, there are 96 samples of neutral, and 192 samples in each of the other categories, a total of 1500 samples, I got 192 x 5 = 960 samples from this dataset. in the SAVEE, there are 60 samples per category, a total of 480 samples, I got 60 x 4=240 samples from this dataset. Because there are about 2,000 samples in total, the number of our data sets should be the same.

Hi, did you solve this problem? I still don't understand what data should be added to improve the accuracy of work.