MediaEval 2019 Emotion and Theme Recognition Task
We are pleased to announce a new MIR-related task held within the MediaEval 2019 evaluation campaign: Emotion and Theme Recognition in Music Using Jamendo.
The Benchmarking Initiative for Multimedia Evaluation (MediaEval) organizes an annual cycle of scientific evaluation tasks in the area of multimedia access and retrieval. In our task, we invite the participants to try their skills at predicting mood and theme tags associated with music recordings using audio analysis and machine learning algorithms.
The task is framed as an auto-tagging problem with tags specific to moods and themes (e.g., happy, dark, epic, melodic, love, film, space). To build the dataset for this task we used a collection of music from Jamendo that is available under the Creative Commons licenses with tag annotations that come from authors/musicians.
All interested researchers are warmly welcomed to participate. See the MediaEval Multimedia Benchmark Workshop for the results of the challenge.