This is a basic demo of a tested architecture from the article "Detection of atypical vehicle trajectories using deep autoencoders in Santiago, Chile"
- Data sets
Due to the private nature of the data sets, we only share a small anonymized sample of the data set to test the neural network code.
This anonymized data set can be obtained from:
https://1drv.ms/u/s!AnkU8l4kGM9wlhupTaHVeHboI36m?e=yjjFZN
- Code
The main code is Demo_SAutoencoder_conv.ipynb which shows the training of a convolutional architecture tested in paper. The dense alternative should be easily implemented.