Table 4 The performance comparison of framework with embedding PAE module and the SOTA in the Public Dataset BS-80K.
Model | Value | Map |
|---|---|---|
Backbone | Mean ± Std | 50.0 ± 2.6% |
Median | 49.5% | |
95% CI | 47.9%, 52.1% | |
Effect sizes | 5.914 | |
p value | *\(p<0.0125\) | |
Cascade R-CNN20 | Mean ± Std | 61.2 ± 0.7% |
Median | 61.2% | |
95% CI | 60.6%, 61.7% | |
Effect sizes | 2.645 | |
p value | *\(p<0.0125\) | |
Faster R-CNN20 | Mean ± Std | 61.9 ± 0.6% |
Median | 61.9% | |
95% CI | 61.4%, 62.4% | |
Effect sizes | 1.914 | |
p value | *\(p<0.0125\) | |
Backbone+DH module | Mean ± Std | 62.0 ± 0.6% |
Median | 62.0% | |
95% CI | 61.5%, 62.5% | |
Effect sizes | 1.23 | |
p value | *\(p<0.0125\) | |
Our SAAI-BMDetector (with PAE module) | Mean ± Std | 63.1 ± 0.7% |
Median | 63.1% | |
95% CI | 62.6%, 63.6% |
