• Proposed a recurrent attention model to handle sequential weakly labeled multi-activity recognition and localization tasks. This work greatly mitigates the tedious annotation work for wearable sensor data.

• The model is trained to repeatedly perform steps of attention on multiple activities of one sample and each step is corresponding to the current focused activity.

• Implemented the whole pipeline for this project. It includes data collection of sequential weakly labeled wearable sensor data, recurrent attention model training, offline model evaluation, and model serving via RESTful API.

K. Wang, J. He, and L. Zhang, “Sequential weakly labeled multiactivity localization and recognition on wearable sensors using recurrent attention networks,” IEEE Transactions on Human-Machine Systems, vol. 51, no. 4, pp. 355–364, 2021.

• Exploited the hard attention model to relief the tedious annotation of training data logged from wearable sensors.

• Built a recurrent hard attention model with CNN backbone for time series sensor data. The trained recurrent attention can focus on the most salient time series parts of a particular activity.

• Whole pipeline was implemented for this project. Including sensor data collection, recurrent hard attention model training, offline model evaluation, and model serving via RESTful API.

J. He, Q. Zhang, L. Wang, and L. Pei, “Weakly supervised human activity recognition from wearable sensors by recurrent attention learning,” IEEE Sensors Journal, vol. 19, no. 6, pp. 2287–2297, 2018.