A machine learning dataset for FRB detection in raw data

We introduce a machine learning (ML) fast radio burst (FRB) dataset that can train ML algorithms to reach the FRBs in raw data. It has 8020 FRB simulation images, 4010 non-FRB and 4010 RFI simulation images, built from public FRB observations and can be expanded to any number as needed. This work provides an open-source dataset for state-of-the-art AI to compare FRB event recognition algorithms. The dataset includes image and Numpy format files for convolutional neural networks (CNN) and classic machine learning algorithms and can implement FRB/non-FRB or FRB/RFI/Blank classification. In the example, we use 31 classical CNN algorithms with pretrained models. In FRB/non-FRB classification, more than 90% accuracy is achieved in the first training epoch, and 99.8% maximum accuracy is achieved in real FRB dataset testing.