Multi-task Deep Learning based Environment and Mobility Detection for User Behavior Modeling

Abstract—Rather than the visual images, the face recognition of the caricatures is far from the performance of the visual images. The challenge is the extreme non-rigid distortions

[7, 8, 9, 10] proposed an unsuper- vised auxiliary task based on a layer-wise training which leads to the concept of pre-training which is a sequential transfer learning consisting

• Train all pedestrian samples also using the pedestrian actions tags (cross/not cross) with the CNN as mentioned above for detection and action recognition based on the Joint

For each category of learning setup, i.e., supervised, semi-supervised, and unsu- pervised, one ML-based and one DL-based technique are evaluated and compared in terms

Real Time Object Detection, Tracking, and Distance and Motion Estimation based on Deep Learning: Application to Smart Mobility.. In Proceedings of the 2019 Eighth

While the best approach obtained an accuracy of 53% in a multi-class classification, the proposed approach allows to detect an abnormal connection with a rate of 84%, and to predict

We then design a multi-task neural-network architecture composed of one joint convolutional network connected to 3 networks ded- icated to the different tasks: ship

Index Terms—AIS, maritime surveillance, deep learning, trans- fer learning, trajectory reconstruction, anomaly detection, vessel type identification, variational recurrent