Table 1 Architecture of the Baseline-conv5 model
From: Object tracking framework with Siamese network and re-detection mechanism
Layer | Kernel Size | Stride | Action Size | Chans. | |
|---|---|---|---|---|---|
For exemplar | For search | ||||
|  |  |  | 127 × 127 | 255 × 255 | 3 |
Conv1 | 11 × 11 | 2 | 59 × 59 | 123 × 123 | 96 |
Pool1 | 3 × 3 | 2 | 29 × 29 | 61 × 61 | 96 |
Conv2 | 5 × 5 | 1 | 25 × 25 | 57 × 57 | 256 |
Pool2 | 3 × 3 | 1 | 23 × 23 | 55 × 55 | 256 |
Conv3 | 3 × 3 | 1 | 21 × 21 | 53 × 53 | 384 |
Conv4 | 3 × 3 | 1 | 19 × 19 | 51 × 51 | 384 |
Conv5 | 3 × 3 | 1 | 17 × 17 | 49 × 49 | 32 |