Tip Recognition

This program uses a lightweight CNN to classify if the region in the tip rack below the pipette is a tip or hole, after moving the pipette vertically above the exact position of the tip/hole.

The accuracy is almost 100%.

Dataset Direction

The dataset should be organized as follows:

|--tip_recognition    
  |--image_classification_CNN  
  |  |--train.py  
  |  |--test.py   
  |  |--dataset   
  |  |  |--train   
  |  |  |  |--hole   
  |  |  |  |  |--iamge_hole_id.jpg   
  |  |  |  |  |--   
  |  |  |  |--tip   
  |  |  |  |  |--image_tip_id.jpg   
  |  |  |--val     
  |  |  |  |--hole   
  |  |  |  |--tip   
  |  |  |--test     
  |  |  |  |--hole   
  |  |  |  |--tip   

Train

cd image_classification_CNN
python train.py

It will plot figures of metrics and output metrics as csv.

Test

Currently please change the dir of images to be tested in test.py according to your real dir.

cd image_classification_CNN
python test.py