Human bone point detection: top-down (part of the theory)

Human bone point detection ： The top-down

1. Datasets

2. Human Pose Estimation

Top Down: The top-down → Find someone first Find some later
Bottom Up: Bottom up → Find some first After induction

3. Keypoints Evaluation Metric

Precision measures how accurate is your predictions. i.e. the percentage of your predictions are
correct.
Recall measures how good you find all the positives. For example, we can find 80% of the
possible positive cases in our top K predictions.

Format of real joint points ：[𝑥1, 𝑦1, 𝑣1, … 𝑥𝑘, 𝑦𝑘, 𝑣𝑘]
Coordinates [𝑥, 𝑦] visible: 𝑣 𝑣 = 0: Unmarked points
𝑣 = 1: Marked but not visible
𝑣 = 2: Marked and image visible
𝑑i Is the Euclidean distance between the annotation and prediction joint points

4. Standard Deviation per Keypoint Type

5. Human Pose Estimation: Top Down

Human target detection :
One Stage: Yolo, SSD etc. ;
Two Stages: Faster RCNN, Mask RCNN etc. ;
Anchor Free: CornerNet, CenterNet;
Key point detection ：

6. FPN：Feature Pyramid Network

C1-C5 The size of the feature map is different ;
Use 1×1 Convolution assurance depth Get the same first P5,
Then up sampling ensures that the size of the characteristic matrix matches ,
Then add the characteristic matrix , Do the same and get P4,P3,P2.