Method of pedestrian activity recognition using limited data and meta-learning

Pedestrian activity recognition is embodied in a method, system, non-transitory computer-readable and vehicle. A Siamese neural network is trained to recognize a plurality of pedestrian activities by training it recordings of the same pedestrian activity from two or more separate training image capture devices. The Siamese neural network is deployed with continual data collection from an additional image capture device to create a dataset of clusters of similar activities in an unsupervised manner. A spatio-temporal intent prediction model is then trained that can be deployed to recognize and predict pedestrian activity. Based on the likelihood of a particular pedestrian activity occurring or currently being underway, an automatic vehicle maneuver can be executed to navigate the situation.