FSDnoisy18k is an open audio dataset. The dataset contains 42.5 hours of audio for 20 sound event classes, including a small amount of manually labeled data and a large amount of real-world noisy data. The audio content is taken from Freesound. The dataset contains a noise set consisting of 15,813 audio clips (38.8 hours) and a test set consisting of 947 audio clips (1.4 hours) with correct labels. The dataset has two main types of label noise: in-vocabulary (IV) and out-of-vocabulary (OOV).