Paste "foa_dev" & "metadata_dev" folders here.
As mentioned from the paper, we used all 1200 synthetic instances as a training set, used 67 "dev-train" instances from STARSS22 to validate the model, and 54 "dev-test" STARSS22 data are exploited at the test phase.
-
[DCASE2022 Task 3] Synthetic SELD mixtures for baseline training
-
STARSS22: Sony-TAu Realistic Spatial Soundscapes 2022 dataset
.
├── foa_dev\ ".wav" audio files should be here
│ ├── dev-test\
│ ├── dev-train\
│ └── dev-val\
│
└── metadata_dev\ ".csv" label files should be here
├── dev-test\
├── dev-train\
└── dev-val\