This page explains how to download the recorded clips. The training + validition sets are downloaded separately from the public evaluation set.<\/p>\n
cd real_data\/src<\/code><\/li>\npython download_real_data.py<\/code><\/li>\n- Send a mail with the csv files in the
real_data\/missing_files<\/code> folder to nicolas<\/a> (and romain<\/a>)<\/li>\n- If you want to do the dcase2019 repo, launch
create_dcase2019_dataset.sh<\/code> from real_data<\/code> folder<\/li>\n<\/ul>\n<\/li>\n<\/ul>\nPublic evaluation set
\n<\/em><\/h4>\nThe evaluation data are in the following repo: DESED_public_eval<\/a><\/strong>.<\/p>\nIt corresponds to “youtube” subset in the desed eval paper<\/a> and in the task 4<\/a> of DCASE 2019 Challenge.<\/p>\n\n- Download DESED_public_eval.tar.gz<\/li>\n
tar -xzvf DESED_public_eval.tar.gz<\/code><\/li>\n- To move it to dcase2019, merge
dataset\/<\/code> with dcase2019\/dataset<\/code>.<\/li>\n<\/ul>\nClass-wise statistics<\/span><\/h4>\n <\/p>\n
\n\n\n<\/td>\n Training (weak)<\/td>\n Validation<\/td>\n Public Evaluation<\/td>\n<\/tr>\n \n<\/td>\n clips<\/td>\n clips<\/td>\n events<\/td>\n clips<\/td>\n events<\/td>\n<\/tr>\n \nAlarm\/bell\/ringing<\/td>\n 205<\/td>\n 187<\/td>\n 420<\/td>\n 79<\/td>\n 196<\/td>\n<\/tr>\n \nBlender<\/td>\n 134<\/td>\n 80<\/td>\n 96<\/td>\n 73<\/td>\n 84<\/td>\n<\/tr>\n \nCat<\/td>\n 173<\/td>\n 121<\/td>\n 341<\/td>\n 70<\/td>\n 240<\/td>\n<\/tr>\n \nDishes<\/td>\n 184<\/td>\n 171<\/td>\n 567<\/td>\n 136<\/td>\n 488<\/td>\n<\/tr>\n \nDog<\/td>\n 214<\/td>\n 160<\/td>\n 570<\/td>\n 82<\/td>\n 441<\/td>\n<\/tr>\n \nElectric shaver\/toothbrush<\/td>\n 103<\/td>\n 62<\/td>\n 65<\/td>\n 84<\/td>\n 108<\/td>\n<\/tr>\n \nFrying<\/td>\n 171<\/td>\n 89<\/td>\n 94<\/td>\n 88<\/td>\n 90<\/td>\n<\/tr>\n \nRunning water<\/td>\n 343<\/td>\n 197<\/td>\n 237<\/td>\n 92<\/td>\n 109<\/td>\n<\/tr>\n \nSpeech<\/td>\n 550<\/td>\n 627<\/td>\n 1754<\/td>\n 314<\/td>\n 913<\/td>\n<\/tr>\n \nVacuum cleaner<\/td>\n 167<\/td>\n 91<\/td>\n 92<\/td>\n 94<\/td>\n 96<\/td>\n<\/tr>\n \nTotal<\/td>\n 1578<\/td>\n 1168<\/td>\n 4093<\/td>\n 692<\/td>\n 2765<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n","protected":false},"excerpt":{"rendered":"This page explains how to download the recorded clips. The training + validition sets are downloaded separately from the public evaluation set. Training set and validation set The real_data folder Clone this repo cd real_data\/src python download_real_data.py Send a mail with the csv files in the real_data\/missing_files folder to nicolas\u2026<\/p>\n