Segmenting and tracking cell instances with cosine embeddings and recurrent hourglass networks.

•CNN based framework for cell instance segmentation and tracking.•Incorporating temporal information as ConvGRUs into a fully convolutional network.•Embedding loss based on cosine similarities to represent individual instances.•Extensive evaluation on muscle fiber and celltracking microscopy images.•State-of-the-art performance on six datasets of the ISBI celltracking challenge.