https://medium.com/@jonathan_hui/map-mean-average-precision-for-object-detection-45c121a31173
For object detection, you should set confidence threshold value which will be used for classification. You can make a graph with precision and recall based on difference threshold values. For example, recall will be close to a very low value and precision will be close to 100% with a high threshold value. on the other hand,it will be opposite with a low threshold value.
Average precision means an average value from each precision with different threshold values. By the way, in a graph above, precision values should have a linear type of decreasing continually. So, a maximum precision value should be applied if the value increase suddenly. (Please, check the url above. )
The mean average precision is an average value with all class objects.
'Papers > Machine learning' 카테고리의 다른 글
| batch, epoch, iteration (0) | 2019.04.15 |
|---|---|
| K-Means clustering (0) | 2019.04.14 |
| How gpu and tpu works (0) | 2019.01.30 |
| High Performance Front Camera ADAS Applications on TI's TDA3X Platform (0) | 2019.01.25 |
| Generative adversarial network (0) | 2019.01.14 |