depressiondectection's Introduction

DepressionDectection

This depression detection work is run on Jupyter notebook.

Dataset

DAIC: https://dcapswoz.ict.usc.edu/
AVID: http://avec2013-db.sspnet.eu/
To obtain both DAIC and AVID datasets, agreement forms should be signed and return to corresponding email address

How to run

DAIC

Classification model training code and regression model training code are provided in classfication folder and regression folder. To run the code, prefix in BiLSTM.ipynb, cnn_audio.py and fusion_net.ipynb should be set to the path where DAIC dataset placed. To run fusion_net.ipynb, path of trained lstm_model and cnn_model should be set corresponding path.

AVID

Preprocessing code of audio recordings in AVID dataset is offered in AudioPreprocess.ipynb. You should change the paths to your AVID dataset path before runing. Regression model training code is provided in cnn_audio_reg_avid.py. The input required by the training code is preprocessed audio clips, which is saved in avid_info.pkl. Therefore, prefix varaible should be set to the path where avid_info.pkl stored during the preprocessing step.

depressiondectection's People

Stargazers

Watchers

depressiondectection's Issues

appeal for help

Hello,I’m a postgraduate student from UCAS(University of Chinese Academy of Sciences).
I'm doing experiment about depression regression,and I need AVEC 2013 dataset.
But it doesn't have download link.
Does it convenient for you to give me this data set?
Thanks!

Looking for the AViD Database

The link for the AViD database does not work anymore. Do you happen to have those files saved locally, and would share them with me?

File Not Found: /Users/apple/Downloads/depression/queries.txt | BiLSTM.ipynb

In the BiLSTM.ipynb file, queries.txt file is mentioned. Is it the Patient Health Questionnaire (PHQ-8) file ?
Could you please help with this by providing the file ?

Data Processing Strategies and Decision-Making in Addressing Data Imbalance

Greetings, esteemed author,

I have observed your utilization of data processing techniques during the phase of data handling for addressing data imbalance. Notably, you opted for oversampling to counterbalance the skewed data distribution. Additionally, in pursuit of alignment, you opted to truncate the initial ten sentences. The rationale behind these actions intrigues me. Is this approach indeed judicious, and what prompts the choice of ten sentences? I find myself perplexed and eager to receive your insightful response.

Recommend Projects