目标检测——day60 Object Detection Using Deep Learning Methods in Traffic Scenarios（待更新）

1 INTRODUCTION

提出了目标检测在交通场景的应用，在介绍深度模型对目标检测带来的提升后，详解目标检测对自动驾驶的安全性等提升，得出结论：目标检测在交通场景中扮演重要的角色。

5 DATASETS AND PERFORMANCE EVALUATION

5.1 Datasets

根据不同目标对象汇总数据集（Table 2）

5.2 Evaluation Criteria

For object detection tasks in traffic scenarios, both detection speed and accuracy are important evaluation criteria.

1. Frame Per Second (FPS)

2. The indicator for measuring the detection accuracy is mean Average Precision (mAP)

3. Intersection over Union (IoU)
   

   Bp represents the detector-predicted box, and Bgt is the ground truth box.When evaluating the prediction box, an IoU threshold such as 0.5 is usually set.

4. True Positive (TP:IoU大于阈值). The opposite is false positive (FP).
   

5. AP，Precision-Recall (PR) curve；The general definition ofAP is to find the area under the PR curve, as shown in Equation (3)
   , as shown in Equation (3)