Purpose:
To check if screw exists in each screw hole on D piece of notebook by Machine Learning.
 |
| Field test of MI4FA project : labeling 3 vacant holes and 1 hole with screw accurately; right pane indicated results of 4 holes with probabilities |
Pictures of Field Test
Training:
Step:
Construct framework to take picture of piece D with conditioned ambient light spotted on
1.Collect pictures of screw hole with and without screw on notebook D piece in every project as many as possible.
2.Auto crop circular region of screw hole in pictures by OpenCV program
3.Train CNN model
training dataset:
Inference:
1.Crop loosely
2.Inverse
3.Threshold
4.pyr
pyrUp
pyrDown
5.can algorithm
6.hough
7.square cropping
8.circle cropping with black padding
9.predict and labeling on picture directly
Result:
留言
張貼留言