Action Recognition on ATRACT custom drone-captured and wearable sensor dataset
85.7AccuracyR(2+1)D
Evaluation Results
| Method | Links | |
|---|---|---|
| R(2+1)DBackbone=R(2+1)D, Sensor Data Augmentation=without (w/o), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 85.7 | |
| R(2+1)DBackbone=R(2+1)D, Sensor Data Augmentation=with (w), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 85.7 | |
| ATRACTBackbone=two-layer CNN, Sensor Data Augmentation=with (w), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 85.7 | |
| S3DBackbone=S3D, Sensor Data Augmentation=without (w/o), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 71.4 | |
| S3DBackbone=S3D, Sensor Data Augmentation=with (w), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 71.4 | |
| R3DBackbone=R3D, Sensor Data Augmentation=with (w), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 71.4 | |
| MC3Backbone=MC3, Sensor Data Augmentation=without (w/o), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 71.4 | |
| MC3Backbone=MC3, Sensor Data Augmentation=with (w), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 71.4 | |
| ATRACTBackbone=two-layer CNN, Sensor Data Augmentation=without (w/o), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 71.4 | |
| R3DBackbone=R3D, Sensor Data Augmentation=without (w/o), Modalities=Video + Sensor (HR + BR + BP + Move.), Fusion Strategy=Late fusion2026.05 | 57.1 |