Map pin anomaly detection

I created this project to participate in the event. By the way, learn openvino.

step 1.

Collect pictures. I sacrificed some pins for this.

Types are divided into: good, broken, crook and short.

good

broke

crook

short

Step 2.

Create a mask image of abnormal pins. Note that the picture format and size need to be the same as the training images.

Check my dataset on github.

Step 3.

Follow anomalibnotebooks to training and test.

Tips

It would be better to remove the background from the images used for training, or use a single color background.

If the number of training times is not enough, the chance of misjudgment will be higher.

Try different modules.