Script for training and testing on Fashion Mnist dataset for Machine learning purpose Based on https://docs.opencv.org/3.4/dc/d88/tutorial_traincascade.html tutorial
Generate training and testing image from fashion mnist byte file
python3 fashionGenerateImage.py
Generate info.dat for positive sample and bg.txt for negative sample with the command labels which allow you to choose what you want to detect
Each training and test example is assigned to one of the following labels:
| Label | Description |
|---|---|
| 0 | T-shirt/top |
| 1 | Trouser |
| 2 | Pullover |
| 3 | Dress |
| 4 | Coat |
| 5 | Sandal |
| 6 | Shirt |
| 7 | Sneaker |
| 8 | Bag |
| 9 | Ankle boot |
python3 fashionGenerateImage.py --labels 0
opencv_createsamples -info info.dat -vec positive.vec -w 28 -h 28 -num 6000
2000 positive image and 2000 negative image are the default value and it took me 2 hours with 10 stages
opencv_traincascade -data training -vec positive.vec -bg bg.txt -numPos 2000 -numNeg 2000 -w 28 -h 28 -numStages 10
This will generate a cascade.xml file in the training folder, so you can now test whether your training has been effective
We have training data, that we have generate before, to test we will use a script which give some information like
- Sensibilité
- Sépcificité
- Précision
- F-mesure
And a file result.png
python3 testCascade.py --labels 0 --cascade /path/to/your/cascade
Default value :
- Labels = 0
- Cascade path = training/cascade.xml