[coursera/ConvolutionalNeuralNetworks/week3]Object Detection(summary&question)

3.1 Object Detection

First, detect the goal whether it's in the image. (yc)

Second, locate it.(dx,dy,dw,dh)

Third, detect what object it is.(c: class)

3.2 Landmark Detection

hand-made feature

3.3 object detection

sliding windows detection

stride: small to large

3.4 convolutional implementation of sliding windows

full connect to convolution

3.5 bounding box predictions

convolutional implementation of sliding windows is not much accurate

YOLO: you only look once algorithm

3.6 Intersection Over Union

3.7 Non-max Suppression

detect each car for one time

highlight the box of biggest pc 

3.8 Anchor Boxes

detect two things in one box

3.9 yolo algorithm

1.localize the object

2.Sliding window detection with convolutional

3.bounding box predictions

4.IoU(Intersection over union)

5.Non-MAX SUPPRESSION

6.anchor box

question:

wrong answer:

3、

4、2N