Table 4 Effect of dilated convolution and grouped convolution on the detection results of LEVIR data set
From: Performance analysis of different DCNN models in remote sensing image object detection
CSP | DC | GC | FASN | mAP@.5 | mAP@[.5:.95] | Time | Memory |
|---|---|---|---|---|---|---|---|
\(\surd\) | \(\times\) | \(\times\) | \(\times\) | 0.882 | 0.639 | 12.3 ms | 420.8 M |
\(\surd\) | \(\surd\) | \(\times\) | \(\times\) | 0.894 | 0.641 | 14.5 ms | 371.1 M |
\(\surd\) | \(\surd\) | \(\surd\) | \(\times\) | 0.892 | 0.642 | 13.7 ms | 212.8 M |
\(\surd\) | \(\surd\) | \(\surd\) | \(\surd\) | 0.908 | 0.667 | 15.4 ms | 291.1 M |