https://urbansounddataset.weebly.com/urbansound8k.html
Basic summary: There are 8732 .wav files of 10 different urban sounds including dog barks, car horns, gun shots (this dataset was taken in Chicago), etc. The dataset is divided in 10 folds (folders) to make the train and test easier. I used fold 1-9 to train the model, and fold 10 to test it. Training it with a powerful GPU, it resulted in an above 90% accuracy, which is pretty good.
This was more of a learning experience. I looked at other's solution to other's problems and employed my own. I read an article (aqibsaeed.github.io/2016-09-03-urban-sound-classification-part-1/) that was extremely helpful.
I'll hopefully learn how reccurent neural networks work in general in the near future. RNN are designed for stuff like audio, video, etc.