image preprocessing
Project description
imagepreprocessing
A small library for speeding up the dataset preparation and model testing steps for deep learning on various frameworks. (mostly for me)
What can it do
- Creates all the required files for darknet-yolo3,4 training including cfg file with default parameters and class calculations in a single line. (example usage)
- Creates train ready data for image classification tasks for keras in a single line.(example usage)
- Makes multiple image prediction process easier with using keras model from both array and directory.
- Predicts and saves multiple images on a directory with using darknet.
- Includes a simple annotation tool for darknet-yolo style annotation. (example usage)
- Auto annotation by given random points for yolo.(example usage)
- Draws bounding boxes of the images from annotation files for preview.
- Plots training history graph from keras history object.(example usage)
- Plots confusion matrix.(example usage)
This dataset structure is required for most of the operations
my_dataset
|----class1
| |---image1.jpg
| |---image2.jpg
| |---image3.jpg
| ...
|----class2
|----class3
...
Install
pip install imagepreprocessing
Create required files for training on darknet-yolo
from imagepreprocessing.darknet_functions import create_training_data_yolo
main_dir = "datasets/my_dataset"
create_training_data_yolo(main_dir)
Create training data for keras
from imagepreprocessing.keras_functions import create_training_data_keras
source_path = "datasets/my_dataset"
save_path = "5000images_on_one_file"
train_x, train_y, valid_x, valid_y = create_training_data_keras(source_path, save_path = save_path, image_size = (299,299), validation_split=0.1, percent_to_use=0.5, grayscale = True)
Make multiple image predictions from directory with keras model
from imagepreprocessing.keras_functions import make_prediction_from_directory_keras
predictions = make_prediction_from_directory_keras("datasets/my_dataset/class1", "models/alexnet.h5")
Create training history graph for keras
from imagepreprocessing.keras_functions import create_history_graph_keras
# training
# history = model.fit(...)
create_history_graph_keras(history)
Make prediction from test array and create the confusion matrix with keras model
from imagepreprocessing.keras_functions import create_training_data_keras, make_prediction_from_array_keras
from imagepreprocessing.utilities import create_confusion_matrix, train_test_split
images_path = "datasets/my_dataset"
# Create training data split the data
x, y, x_val, y_val = create_training_data_keras(images_path, save_path = None, validation_split=0.2, percent_to_use=0.5)
# split training data
x, y, test_x, test_y = train_test_split(x,y,save_path = save_path)
# ...
# training
# ...
class_names = ["apple", "melon", "orange"]
# make prediction
predictions = make_prediction_from_array_keras(test_x, model, print_output=False)
# create confusion matrix
create_confusion_matrix(predictions, test_y, class_names=class_names, one_hot=True)
create_confusion_matrix(predictions, test_y, class_names=class_names, one_hot=True, cmap_color="Blues")
Make multi input model prediction and create the confusion matrix
from imagepreprocessing.keras_functions import create_training_data_kera
from imagepreprocessing.utilities import create_confusion_matrix, train_test_split
import numpy as np
# Create training data split the data and split the data
source_path = "datasets/my_dataset"
x, y = create_training_data_keras(source_path, image_size=(28,28), validation_split=0, percent_to_use=1, grayscale=True, convert_array_and_reshape=False)
x, y, test_x, test_y = train_test_split(x,y)
# prepare the data for multi input training and testing
x1 = np.array(x).reshape(-1,28,28,1)
x2 = np.array(x).reshape(-1,28,28)
y = np.array(y)
x = [x1, x2]
test_x1 = np.array(test_x).reshape(-1,28,28,1)
test_x2 = np.array(test_x).reshape(-1,28,28)
test_y = np.array(test_y)
test_x = [test_x1, test_x2]
# ...
# training
# ...
# make prediction
predictions = make_prediction_from_array_keras(test_x, model, print_output=False, model_summary=False, show_images=False)
# create confusion matrix
create_confusion_matrix(predictions, test_y, class_names=["0","1","2","3","4","5","6","7","8","9"], one_hot=True)
Create required files for training on darknet-yolo and auto annotate images by center
Auto annotation is for testing the dataset or just for using it for classification, detection won't work without proper annotations.
from imagepreprocessing.darknet_functions import create_training_data_yolo, auto_annotation_by_random_points
import os
main_dir = "datasets/my_dataset"
# auto annotating all images by their center points (x,y,w,h)
folders = sorted(os.listdir(main_dir))
for index, folder in enumerate(folders):
auto_annotation_by_random_points(os.path.join(main_dir, folder), index, annotation_points=((0.5,0.5), (0.5,0.5), (1.0,1.0), (1.0,1.0)))
# creating required files
create_training_data_yolo(main_dir)
Annotation tool for derknet-yolo
from imagepreprocessing.darknet_functions import yolo_annotation_tool
yolo_annotation_tool("test_stuff/images", "test_stuff/obj.names")
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
imagepreprocessing-1.6.0.tar.gz
(17.6 kB
view hashes)