Learning Activity Progression in LSTMs for Activity Detection and Early Detection

TLDR

Traditional LSTM training for activity recognition uses only classification loss, ignoring temporal progression. This study aims to improve temporal deep model training by enforcing that detection scores for the correct activity monotonically increase as more of the activity is observed, thereby enhancing activity detection and early detection. The authors introduce ranking losses that penalize violations of this monotonicity, combining them with standard classification loss to train LSTM models. On ActivityNet, the proposed ranking losses yield significant improvements in both activity detection and early detection performance.

Abstract

In this work we improve training of temporal deep models to better learn activity progression for activity detection and early detection tasks. Conventionally, when training a Recurrent Neural Network, specifically a Long Short Term Memory (LSTM) model, the training loss only considers classification error. However, we argue that the detection score of the correct activity category, or the detection score margin between the correct and incorrect categories, should be monotonically non-decreasing as the model observes more of the activity. We design novel ranking losses that directly penalize the model on violation of such monotonicities, which are used together with classification loss in training of LSTM models. Evaluation on ActivityNet shows significant benefits of the proposed ranking losses in both activity detection and early detection tasks.