Article Source

• Convolutional LSTMs for Video Prediction

Convolutional LSTMs for Video Prediction

• London TensorFlow Meetup
• Using Convolutional LSTMs for video prediction. Applications to self-driving cars and medical image processing (Armando Vieira, Data Scientist)

Abstract

Convolutional LSTM are a class of recurrent network with Long Short Term Memory (LSTM) units applied over convolutional networks (CNN). They are particularly useful to for unsupervised videos analysis, either image segmentation, classification of annotation. We will present how to create a convolutional LSTM model in Keras and Tensorflow for image segmentation and show some results for self-driving cars and for annotation of arteries and veins on ultra-sound videos.