Table 2 Reduction of MR by combining our deep motion feature with various appearance features
From: Pedestrian detection with motion features via two-stream ConvNets
Base method | ACF-ours | LDCF-ours | CCF | Two-stage CNN |
|---|---|---|---|---|
Appearance feature only | 36.7 | 33.4 | 18.8 | 13.6 |
With our deep motion feature | 31.4 | 30.6 | 16.4 | 12.4 |