The 16th Winter School on Audio and Speech Processing 2023 is pleased to announce two challenges for the WiSSAP Cup 2023. More information about each challenge can be found by visiting the challenge page from the links below. The top-3 ranked teams of each challenge will be invited to join the WiSSAP 2023 program in person from 18th – 21st December. The winners of both challenges shall be felicitated at the closing ceremony of the school.
Prof. Vipul Arora will also give 3 tutorials discussing both challenges. RSVP now!
Finding the same or similar audio in a massive audio corpus is computationally and memory expensive.
Audio fingerprinting is a technique to index audio datasets for efficient retrieval. Retrieval entails matching the database entries with a given query audio.
The task in this challenge is to develop method to index a large audio database and search a given audio query efficiently. The query could be a noisy and distorted segment from one of the audios in the database.
Query-by-humming means to search for a song by humming its tune.
A general approach is to extract pitch information from each audio in the database and match it efficiently with the pitch information extracted from the query.
This task is challenging due to several factors, including the variability of input signals from a person humming or singing, differences in tempo between recordings and queries, and background instruments/noise in the pieces being matched.
The algorithms are evaluated using measures such as hit rate.
- Saturday 12th August 2023 - Registration of Teams begins (Link- here)
- 5th September 2023 - Release Development Dataset
- 30th September 2023 - Registration of Teams closes
- 30th October 2023 - Test Set Release
- 3rd November 2023 - Final Submission of Test Results and 1-page report on the method used
- 6th November 2023 - Results Announced and Invitations to Top Teams
- 18th December 2023 - 21th December 2023 - WiSSAP 2023
- Registration of teams for either/both of the two challenges starts on August 12.
- Each team must have no more than 2 participants.
- The competitions will take place on Kaggle. Kindly ensure to sign up on Kaggle before registration.
- After the release of the development dataset and/or baseline on September 5, the teams can start working on the problem. Participants are still allowed to register their teams until September 30.
- The end of the development phase and release of test set will be on October 30th.
- Every team is expected to record their final results (test set inference, etc.) and submit them along with a 1-page manuscript outlining the method developed.
- There are restrictions on the scale/size of the method/model, as mentioned in specific task guidelines.
- We have planned 3 tutorials discussing both challenges, coving topics including Background and Applications of Music Search, Existing Works in the area and much more. RSVP’d participants will receive invite to the talks. RSVP now! The schedule is as follows-
- 12th September - Background and Introduction (Talk here. Slides here.)
- 2nd October - Mathematical Foundations
- 15th October - Ask the Experts
- Submissions adhering to all competition policies shall be ranked based on a task-specific metric, and the top 3 teams shall be invited for felicitation during the closing ceremony of the winter-school program.
The teams are first required to register themselves by clicking the link below. The submission guidelines for each of the challenges are mentioned on their individual pages.
Frequently Asked Questions
We will fill this section with questions we receive and our answers.
If this section does not answer all of your queries, please ask us firstname.lastname@example.org.