added image eda app section

app.py

@@ -29,7 +29,8 @@
     selected_structure = st.selectbox("Choose data structure to analyse", 
                                       ("Structured data", 
                                        "Unstructured, text data",
-                                       "Unstructured, audio data"))
+                                       "Unstructured, audio data",
+                                       "Unstructured, image data"))
 
     if selected_structure == "Structured data":
 
@@ -41,12 +42,16 @@
         from text_eda.text_data import *
         text_data_app()
 
-
     elif selected_structure == "Unstructured, audio data":
 
         from audio_eda.audio_data import *
         audio_data_app()
 
+    elif selected_structure == "Unstructured, image data":
+
+        from image_eda.image_data import *
+        image_data_app()
+
 
 except KeyError:
     st.error("Please select a key value from the dropdown to continue.")

helper_functions.py

@@ -6,6 +6,12 @@
 from PIL import Image
 import pandas as pd
 
+import io
+from PIL import Image
+from pprint import pprint
+from zipfile import ZipFile
+from image_eda.augment import apply_augmentations
+
 def set_bg_hack(main_bg):
     '''
     A function to unpack an image from root folder and set as bg.
@@ -173,4 +179,67 @@ def check_input_method(data_input_mthd):
             st.session_state.is_batch_process = True
             st.session_state.is_file_uploaded = True
 
-    return df,st.session_state.txt
+    return df,st.session_state.txt
+
+def load_images():
+
+    data = st.sidebar.file_uploader("Upload image dataset", 
+                                    type=['png', 'jpg', 'jpeg'],
+                                    accept_multiple_files=True) 
+
+    if data:
+        images = []
+        augmentations = get_augmentations()
+        for image_file in data:
+            file_details = {"None": None, "File name":image_file.name, "File type":image_file.type, "File size":image_file.size}
+            image = Image.open(image_file)
+            images.append((file_details, image))
+
+        images = apply_augmentations(images, augmentations)
+        return images
+
+def _get_default_augmentations() -> dict:
+    augmentations = {
+        'resize': {
+            'width': None,
+            'height': None
+        },
+        'grayscale': False,
+        'contrast': {'value':None},
+        'brightness': {'value':None},
+        'sharpness': {'value':None},
+        'color': {'value':None},
+        'denoise': False,
+    }
+    return augmentations
+
+def get_augmentations() -> dict:
+    if 'augmentations' not in st.session_state:
+        st.session_state.augmentations = _get_default_augmentations()
+    return st.session_state.augmentations
+
+def update_augmentations(augmentations) -> None:
+    st.session_state.augmentations = augmentations
+
+def _file_process_in_memory(images):
+    """ Converts PIL image objects into BytesIO in-memory bytes buffers. """
+    new_images = []
+    for image_name, pil_image in images:
+        file_object = io.BytesIO()
+        pil_image.save(file_object, "PNG")
+        pil_image.close()
+        new_images.append((image_name, file_object))
+
+    return new_images
+
+def export(images):
+    images = _file_process_in_memory(images)
+
+    # Create an in-memory zip file from the in-memory image file data.
+    zip_file_bytes_io = io.BytesIO()
+
+    with ZipFile(zip_file_bytes_io, 'w') as zip_file:
+        for image_name, bytes_stream in images:
+            zip_file.writestr(image_name["File name"]+".png", bytes_stream.getvalue())
+        name = st.sidebar.text_input("File name", value="My augmented dataset")
+        st.sidebar.download_button('Download Zip', zip_file_bytes_io.getvalue(), file_name=f'{name}.zip')

image_eda/augment.py

@@ -0,0 +1,28 @@
+import streamlit as st
+from PIL import Image, ImageEnhance
+import cv2
+import numpy as np
+
+def apply_augmentations(images, augmentations) -> list:
+    new_images = []
+    for details, image in images:
+        new_image = image
+        if augmentations["resize"]["width"] is not None and augmentations["resize"]["height"] is not None:
+            new_image =  new_image.resize((augmentations["resize"]["width"], augmentations["resize"]["height"]), Image.ANTIALIAS)
+        if augmentations["grayscale"] == True:
+            new_image = new_image.convert('L')
+        if augmentations["contrast"]["value"] is not None:
+            new_image = ImageEnhance.Contrast(new_image).enhance(augmentations["contrast"]["value"])
+        if augmentations["brightness"]["value"] is not None:
+            new_image = ImageEnhance.Brightness(new_image).enhance(augmentations["brightness"]["value"])
+        if augmentations["sharpness"]["value"] is not None:
+            new_image = ImageEnhance.Sharpness(new_image).enhance(augmentations["sharpness"]["value"])
+        if augmentations["color"]["value"] is not None:
+            new_image = ImageEnhance.Color(new_image).enhance(augmentations["color"]["value"])
+        if augmentations["denoise"] == True:
+            if len(new_image.split()) != 3:
+                new_image = Image.fromarray(cv2.fastNlMeansDenoising(np.array(new_image)))
+            else:
+                new_image = Image.fromarray(cv2.fastNlMeansDenoisingColored(np.array(new_image)))
+        new_images.append((details, new_image))
+    return new_images

image_eda/image_data.py

@@ -0,0 +1,73 @@
+"""
+
+A file with the high level image eda app functions
+
+"""
+import streamlit as st
+from helper_functions import *
+from image_eda.preparation import show_grid, show_sizes, show_histograms, show_channels, augmentations, export_images
+
+
+def image_data_app():
+
+    st.write("Welcome to the DQW for unsupervised image data. ",
+            "Understanding your data is an important step ",
+            "in AI model development. This app ",
+            "offers visualisation of descriptive statistics of a ",
+            "collection of images using a multitude of packages.")
+
+    display_app_header(main_txt = "Step 1",
+                       sub_txt= "Upload data",
+                       is_sidebar=True)
+
+    images = load_images()
+
+    if images:
+
+        display_app_header(main_txt = "Step 2",
+                        sub_txt= "Choose what you want to see/do",
+                        is_sidebar=True)
+
+        selected_structure = st.sidebar.radio(
+                            "", 
+                            ("Image grid", 
+                            "Image sizes",
+                            "Color histograms",
+                            "Color channels",
+                            "Augmentations"))
+
+        if selected_structure == "Image grid":
+            show_grid(images)
+        elif selected_structure == "Image sizes":
+            show_sizes(images)
+        elif selected_structure == "Color histograms":
+            show_histograms(images)
+        elif selected_structure == "Color channels":
+            show_channels(images)
+        elif selected_structure == "Augmentations":
+            augmentations(images)
+
+            display_app_header(main_txt = "Step 3",
+                            sub_txt= "Export augmented data",
+                            is_sidebar=True)
+
+            is_export = st.sidebar.button("Export ⬇️")
+            if is_export:
+                export_images(images)
+
+
+
+    # prompt the user with an option to select which data they want to 
+    # analyse - save code, right now we only have 1 option
+    #selected_structure = st.selectbox("Choose data structure to analyse", 
+                                       #("Unsupervised image data", 
+                                       #"Supervised image data, Classification",
+                                       #"Supervised image data, Segmentation",
+                                       #"Supervised image data, Regression"))
+
+    #elif selected_structure == "Supervised image data, Classification":
+        #st.error("This feature has not been implemented yet!")
+    #elif selected_structure == "Supervised image data, Segmentation":
+        #st.error("This feature has not been implemented yet!")
+    #elif selected_structure == "Supervised image data, Regression":
+        #st.error("This feature has not been implemented yet!")