Skip to content

README.md

Latest commit

 

History

History
126 lines (96 loc) · 3.71 KB

README.md

File metadata and controls

126 lines (96 loc) · 3.71 KB
Raw

Sample Code

Image Datasets:

Filters and augmentation:

# Import flow2ml package
from flow2ml import Flow

# Give the Dataset and Data directories
flow = Flow( 'dataset_dir' , 'data_dir' )

# Define The Filters to be used
filters = ["median", "laplacian", "gaussian", "sobelx", "sobely", "bilateral"]

# Apply The Filters
flow.applyFilters( filters )

# Define The augmentation operations to be used
   
operations = {'flipped': 'horizontal', 'rotated': 90, 'sheared': {'x_axis': 5, 'y_axis': 15}, 'cropped': [50, 100, 50, 100], 'scaled': 0.1, 'zoomed': 2, 'histogramequalised':True, 'greyscale': True, 'CLAHE':True, 'inverted':True, 'eroded':True, 'dilated':True, 'opened':True, 'closed':True,'thresholded':{'type':'adaptive','thresh_val':0}, 'colorspace':{'input':'BGR','output':'BGR'}, 'canny':{'threshold_1':100,'threshold_2':200}, 'brightnessenhanced':True}

# Apply The Augmentation
flow.applyAugmentation( operations )

# To apply augmentations on filtered images, supply a list of filters to apply the augmentations to using the apply_on_filters key. The below code applies flipping and rotation augmentation on median filtered images.
filters = ["median", "laplacian"]
flow.applyFilters( filters )
operations = {'flipped': 'horizontal', 'rotated': 90, 'apply_on_filters': ['median']}
flow.applyAugmentation( operations )

# To apply filters on augmented images, supply a list of augmentations to apply the filters to using the apply_on_augmentation key. The below code applies median and laplacian filters on the flipped images.
operations = {'flipped': 'horizontal', 'rotated': 90}
flow.applyAugmentation( operations )
filters = ["median", "laplacian", {'apply_on_augmentation': ['flipped']}]
flow.applyFilters( filters )

Train and test splits:

# To be given input by the user.
img_dimensions = (150, 150, 3)

# If working with greyscale image, change the number of channels from 3 to 1.
test_val_split = 0.25

# Select a random seed for splitting the data
random_state = 22

# Select one-hot or label encoding technique
encoding = 'label'

# Obtain Train, Validation data splits
(train_x, train_y, val_x, val_y) = flow.getDataset( img_dimensions, test_val_split, random_state, encoding )

Assesing image quality:

# Create an image quality report using Entropy or BRISQUE for all images
image_quality = "entropy"
flow.calculateImageQuality( image_quality )

Evaluating models:

# For Pytorch and scikit-learn models
from flow2ml import Auto_Results

# Set the Input Model by replacing None
model = None

x = Auto_Results(model, val_x, val_y)

# Call the get_results_docx() function to get the results in a Results folder 
x.get_results_docx()

# For Tensorflow models 
from flow2ml import Tf_Results

# Set the Input Model by replacing None
model = None
x = Tf_Results(model, validation_generator)
x.tf_get_results_docx()

Model conversions:

# Define The conversions to be used
conversions = {'tfjs':True, 'tflite':True}

# convert tensorflow model to tfjs/tflite
flow.deployTensorflowModels( conversions, model )

Please try to maintain the dataset in the following manner in order to run the code easily.

dataset_dir
├──data_dir/
|       ├──Label 1 Folder
|       ├──Label 2 Folder
|       ├──Label 3 Folder  
|               .
|               .
|               .        
|       └──Label n Folder 
| 
└────Other Files

Non image datasets:

Working with .csv files:

# Import the required module
from flow2ml import Process_Csv
import pandas as pd

# Read a csv file
df = pd.read_csv('./Tips.csv')

# Pass the dataframe to the module to generate analysis reports
x = Process_Csv(df)
x.create_analysis_docx()
x.create_visualisation_docx()