Inspired by link
This wheat quality detection test we are using to identify the quality of given wheat grains image. The dataset we used is provided by link.
The wheat quality detection problem is divided into two sub problems given as following:
- Two classs classification i.e. a healthy grain or other.
- Five class slassification i.e. a healthy grain, damaged grain, foreign partical, broken grain and grain cover.
The dataset we used for training (is the single grain or other images) extracted for above mentioned dataset.
- opencv-python
- keras
- tensorflow
- matplotlib
Tested with python3.5
For two class classification:
$ python classifier_2_v2.py
68/68 [==============================] - 0s 499us/step - accuracy: 0.9020 - loss: 0.2475
MLP Test loss: 0.247524231672287
MLP Test accuracy: 0.9019879698753357
For five class classification:
$ python classifier_5_v2.py
65/65 [==============================] - 0s 532us/step - loss: 0.4837 - accuracy: 0.8254
MLP Test loss: 0.483661413192749
MLP Test accuracy: 0.8253890872001648
For performing a saimple test:
$ python cmd_wheat_quality_detector_v2.py
Enter the file(wheat image) location to dectect : test_2.jpg
Segmentation in process...
Level 1 segmentation Finished:
Rejected segment: 1
Level 2 segmentation Finished:
Rejected segment: 21
Total number of segments 124
Number of rejected segments 22
Segmentation in Complete.
Feature extraction in process...
Feature extraction in complete.
Number of good grain : 84
Number Not good grain or imputity: 18
Please feel free to email & contact me if you run into issues or just would like to talk about the future usage.