Mitigating and evaluating static bias of action representations in the background and the foreground
In video action recognition, shortcut static features can interfere with the learning of motion
features, resulting in poor out-of-distribution (OOD) generalization. The video background is …
features, resulting in poor out-of-distribution (OOD) generalization. The video background is …
A knowledge-based hierarchical causal inference network for video action recognition
Currently, existing action recognition methods mainly use a data-driven method to extract
spatio-temporal representations of actions for recognition. However, this method may face …
spatio-temporal representations of actions for recognition. However, this method may face …
A Tutorial on Dataset Creation for Sensor-based Human Activity Recognition
Well-documented, large and diverse datasets are important to facilitate supervised and
transferable deep learning methods for human activity recognition. The past years have …
transferable deep learning methods for human activity recognition. The past years have …
[PDF][PDF] Supplementary Material: Mitigating and Evaluating Static Bias of Action Representations in the Background and the Foreground
No 25.25 72.83 44.65 81.94 36.58 85.03 Mixup 27.64 74.48 47.16 81.85 36.62 82.06
VideoMix 29.37 72.50 42.59 72.21 32.68 76.07 SDN 27.14 71.14 48.47 83.83 34.87 81.88 …
VideoMix 29.37 72.50 42.59 72.21 32.68 76.07 SDN 27.14 71.14 48.47 83.83 34.87 81.88 …