Table 7 Comparison of pooling strategies for low- and high-level feature aggregation.
From: A CNN-RNN Siamese framework with multi-level aggregation for video-based person re-identification
Low-level feature pooling | High-level feature pooling | iLIDS-VID CMC Rank 1 | PRID-2011 CMC Rank 1 |
|---|---|---|---|
Max pooling | Max pooling | 46% | 56% |
Mean pooling | Max pooling | 41% | 66% |
Max pooling | Mean pooling | 54% | 66% |
Mean pooling | Mean pooling | 57% | 76% |
