Research Article
Detection of Pine Cones in Natural Environment Using Improved YOLOv4 Deep Learning Algorithm
Table 1
Structure of the original backbone network and the improved backbone network.
| | Module | Number of layers | Number of filters | Output channels | | Original | Improved | Original | Improved |
| | 1 | Residual∗1 | Residual∗1 | 1 × 1 Conv∗32 | 1 × 1 Conv∗32 | 64 | | 3 × 3 Conv∗64 | 3 × 3 Conv∗64 |
| | 2 | Residual∗2 | Residual∗2 | 1 × 1 Conv∗64 | 1 × 1 Conv∗64 | 64 | | 3 × 3 Conv∗64 | 3 × 3 Conv∗64 |
| | 3 | Residual∗8 | Dense∗4 | 1 × 1 Conv∗128 | 1 × 1 Conv∗64 | 128 | | 3 × 3 Conv∗128 | 3 × 3 Conv∗32 |
| | 4 | Residual∗8 | Dense∗4 | 1 × 1 Conv∗256 | 1 × 1 Conv∗128 | 256 | | 3 × 3 Conv∗256 | 3 × 3 Conv∗64 |
| | 5 | Residual∗4 | Dense∗4 | 1 × 1 Conv∗512 | 1 × 1 Conv∗256 | 512 | | 3 × 3 Conv∗512 | 3 × 3 Conv∗128 |
|
|