Skip to main content

Table 4 Ablative study on UT-Interaction dataset (%)

From: Human activity prediction using saliency-aware motion enhancement and weighted LSTM network

Settings SME Weighted module Accuracy with half videos Accuracy with full videos
Setting 1 x x 81.0 89.6
Setting 2 x 88.5 92.1
Setting 3 x 81.6 90.2
Setting 4 92.6 96.7