DiCOVA 2021 Challenge
First phase
EvaluationEnd
Competition EndsDiCOVA Challenge 2021
Diagnosing COVID-19 using Acoustics
Track-1 (Cough Sound)
Welcome to Track-1 of the DiCOVA challenge!
What is it about?
The COVID-19 pandemic has resulted in more than 100 million infections, and more than 2 million casualties. The global crisis spans across 200 countries. The DiCOVA Challenge is designed to find scientific and engineering insights to the question - Can COVID-19 be detected from the cough sound signals of an individual?. The challenge provides the participants a dataset of sound signals gathered from COVID-19 positive and non-COVID-19 individuals.
Participants are encouraged to analyze the dataset, build classification models and submit their performance scores to a leaderboard. Participants are also encouraged to document their findings as a manuscript and submit to the DiCOVA Special Session at Interspeech 2021 for peer-review. Selected findings will be invited for presentation at the Interspeech 2021, the flagship conference of the global speech science and technology community, to be held in Brno from Aug 31-Sept 3, 2021.
How to participate?
Go to the website: https://dicova2021.github.io/#register and submit the registration form. We will subsequently get in touch with you and send the dataset to your email address. Create a CodaLab account for your Team using your (same as above) email address. That’s it! You are all set to build your classifier models, and submit your performance scores to the leaderboard. If you have any other queries, please contact the organisers at dicova2021@gmail.com.
Evaluation Criteria
Evaluation method:
The participating teams are ranked based on their system performance on the evaluation (blind test) set. The leaderboard will feature the system performance on evaluation (blind test) and the average performance on 5 validation folds provided as part of the development dataset. The score files that participants are required to submit should indicate the probability of COVID positive, for each test recording.
Evaluation metric:
The main metric for the challenge is the “Area under the ROC curve” (AUC).
Receiver operating characteristic (ROC): It is a graphical representation of the true positive rate (TPR) plotted against the false positive rate (FPR) computed at different thresholds of a binary classifier. The thresholds are chosen with a granularity of 0.0001 in the scoring implementation.
True positive rate(TPR) = number of correctly identified positive samples/number of positive samples
TPR is also called as “Sensitivity”
False positive rate(FPR) = number of negative samples wrongly identified as positive / number of negative samples
FPR equals one minus the true negative rate (TNR). TNR is also called the "Specificity".
An illustration is provided below:
Terms and Conditions
Participation in this challenge is open to all who are interested. If you have reached this competition page, we assume you have registered for the challenge at https://dicova2021.github.io/#register and will adhere to the signed Terms and Conditions. Below are a few instructions to be followed for the challenge
- Make sure the “email-ID” used to login into Codalab is the same as that given in the Terms_and_Conditions document
- After creating the Codalab account, go to settings and fill in the “Team Name” as same as that given in the Terms_and_Conditions document
- A team should use only one Codalab account to submit scores to the leaderboard.
Evaluation
Start: March 1, 2021, midnight
Description: Evaluate the blind test and the 5 validation folds scores
Competition Ends
March 23, 2021, noon
You must be logged in to participate in competitions.
Sign In