Implemented new features: first issue and second issue

app/forms/classification_form_histogram.py

@@ -0,0 +1,13 @@
+from flask_wtf import FlaskForm
+from wtforms import SubmitField, SelectField
+from wtforms.validators import DataRequired
+
+from app.utils.list_images import list_images
+from config import Configuration
+
+conf = Configuration()
+
+
+class ClassificationFormHistogram(FlaskForm):
+    image = SelectField('image', choices=list_images(), validators=[DataRequired()])
+    submit = SubmitField('Submit')

app/forms/transformation_form.py

@@ -0,0 +1,20 @@
+from flask_wtf import FlaskForm
+from wtforms import SubmitField, SelectField, FloatField
+from wtforms.validators import DataRequired
+from wtforms.widgets import NumberInput
+
+from app.utils.list_images import list_images
+from config import Configuration
+
+conf = Configuration()
+
+
+class TransformationForm(FlaskForm):
+
+    color = FloatField(widget=NumberInput(step=0.1, min=-10.0, max=10.0), default=1.0, validators=[DataRequired()])
+    brightness = FloatField(widget=NumberInput(step=0.1, min=-10.0, max=10.0), default=1.0, validators=[DataRequired()])
+    contrast = FloatField(widget=NumberInput(step=0.1, min=-10.0, max=10.0), default=1.0, validators=[DataRequired()])
+    sharpness = FloatField(widget=NumberInput(step=0.1, min=-10.0, max=10.0), default=1.0, validators=[DataRequired()])
+
+    image = SelectField('image', choices=list_images(), validators=[DataRequired()])
+    submit = SubmitField('Submit')

app/routes/__init__.py

@@ -1,4 +1,5 @@
 from .classifications import *
+from .transformations import *
 from .classifications_id import *
 from .info import *
 from .home import *

app/routes/classifications.py

@@ -5,7 +5,8 @@
 
 from app import app
 from app.forms.classification_form import ClassificationForm
-from ml.classification_utils import classify_image
+from app.forms.classification_form_histogram import ClassificationFormHistogram
+from ml.classification_utils import classify_image, plot_histogram, clear_dir, plot_histogram_rgb
 from config import Configuration
 
 config = Configuration()
@@ -20,7 +21,6 @@ def classifications():
     if form.validate_on_submit():  # POST
         image_id = form.image.data
         model_id = form.model.data
-
         redis_url = Configuration.REDIS_URL
         redis_conn = redis.from_url(redis_url)
         with Connection(redis_conn):
@@ -38,3 +38,30 @@ def classifications():
     # otherwise, it is a get request and should return the
     # image and model selector
     return render_template('classification_select.html', form=form)
+
+
+@app.route('/classifications_histogram', methods=['GET', 'POST'])
+def classifications_histogram():
+    """API for selecting an image and return the histogram
+    of the image"""
+    form = ClassificationFormHistogram()
+    if form.validate_on_submit():  # POST
+        image_id = form.image.data
+
+        img_path = f'app/static/imagenet_subset/{image_id}'
+        histogram_img_path = f'app/static/imagenet_histogram/hist_{image_id}'
+
+        plot_histogram(img_path, histogram_img_path)
+        # plot_histogram_rgb(img_path, histogram_img_path)
+
+        # histogram viewer with the corresponding image feedback
+        return render_template('histogram_output.html', image_id=image_id)
+    # image selector
+    return render_template('histogram_template.html', form=form)
+
+
+
+
+
+
+

app/routes/home.py

@@ -9,5 +9,7 @@
 @app.route('/home', methods=['GET'])
 @app.route('/index', methods=['GET'])
 @app.route('/', methods=['GET'])
+
 def home():
     return render_template('home.html')
+

app/routes/transformations.py

@@ -0,0 +1,37 @@
+from flask import render_template
+from app import app
+from app.forms.transformation_form import TransformationForm
+from ml.classification_utils import transformation_image
+from config import Configuration
+import uuid
+
+config = Configuration()
+
+
+@app.route('/transformations', methods=['GET', 'POST'])
+def transformations():
+    """API for selecting an image and running a
+    transformation. Returns the enhanced image saving it by unique ID."""
+    form = TransformationForm()
+    if form.validate_on_submit():  # POST
+        image_id = form.image.data
+
+        unique_id = str(uuid.uuid4())
+
+        img_path = f'app/static/imagenet_subset/{image_id}'
+
+        transformation_img_path = f'app/static/imagenet_transform/{unique_id}.JPEG'
+
+        color_id = form.color.data
+        brightness_id = form.brightness.data
+        contrast_id = form.contrast.data
+        sharpness_id = form.sharpness.data
+
+        transformation_image(img_path, transformation_img_path, color_factor=color_id, brightness_factor=brightness_id,
+                             contrast_factor=contrast_id,
+                             sharpness_factor=sharpness_id)
+        # transformation output
+        return render_template("transformation_output.html", image_id=image_id, unique_id=unique_id)
+
+    # image and transformation selector
+    return render_template('transformation_select.html', form=form)

app/templates/histogram_output.html

@@ -0,0 +1,39 @@
+{% extends "base.html" %}
+
+{% block content %}
+
+    <style>
+        .large-front-thumbnail {
+            position: relative;
+            max-width: 100%;
+            height: auto;
+            display: block;
+            margin: 0 auto;
+        }
+
+    </style>
+    <div class="row">
+        <div class="col">
+            <div class="card">
+                <h4>
+                    <th scope="col">Feedback Image</th>
+                </h4>
+                    <img class="large-front-thumbnail"
+                         src="{{ url_for('static', filename='imagenet_subset/'+image_id) }} "/>
+            </div>
+        </div>
+        <div class="col">
+            <div class="card ">
+                <h4>
+                    <th scope="col">Histogram</th>
+                </h4>
+                <table class="table">
+                    <img class="large-front-thumbnail"
+                         src="{{ url_for('static', filename='imagenet_histogram/hist_'+image_id) }} "/>
+                </table>
+                <a class="btn btn-dark mb-2" href="/classifications_histogram" role="button">Back</a>
+            </div>
+        </div>
+    </div>
+{% endblock %}
+

app/templates/histogram_template.html

@@ -0,0 +1,16 @@
+{% extends "base.html" %}
+
+{% block content %}
+
+    <h1>Select Image</h1>
+    <form action="" method="post" novalidate>
+        {{ form.hidden_tag() }}
+        <h4>
+            Image:
+        </h4>
+        <p>
+            {{ form.image }}
+        </p>
+        <button type="submit" class="btn btn-dark mb-2">Submit</button>
+    </form>
+{% endblock %}

app/templates/home.html

@@ -4,9 +4,13 @@
 
     <p style="margin-top:1cm;">This service provides the output scores for popular deep networks
         from the <a href="https://pytorch.org/docs/stable/torchvision/models.html">
-            Torchvision model zoo</a>. Go
-        <a href={{ url_for("classifications") }}>here</a> for
-        trying out the service.</p>
+            Torchvision model zoo</a>. </p>
+    <p> Go <a href={{ url_for("classifications") }}>here</a> for trying out the service.</p>
+
+    <p> Go <a href={{ url_for("transformations") }}>here</a> for trying out the trying out the transformation service.</p>
+
+    <p> Go <a href={{ url_for("classifications_histogram") }}>here</a> if you want to show the histogram for a specific image</p>
+
     <p style="margin-top:1cm;">For the list of models and images
         available, click
         <a href={{ url_for("info") }}>here</a>.</p>

app/templates/transformation_output.html

@@ -0,0 +1,48 @@
+{% extends "base.html" %}
+
+{% block content %}
+
+    <style>
+        .large-front-thumbnail {
+            position: relative;
+            max-width: 100%;
+            height: auto;
+            display: block;
+            margin: 0 auto;
+        }
+
+    </style>
+    <div class="row">
+        <div class="col">
+            <div class="card">
+                <thead>
+                    <h3>
+                        <th scope="col">Original Image</th>
+                    </h3>
+                </thead>
+                <img class="large-front-thumbnail"
+                     src="{{ url_for('static', filename='imagenet_subset/'+image_id) }} "
+                     alt={{ image_id }}/>
+            </div>
+        </div>
+        <div class="col">
+            <div class="card ">
+                <thead>
+                    <h3>
+                        <th scope="col">Transformation</th>
+                    </h3>
+                </thead>
+                <table class="table">
+                    <div class="col">
+                        <div class="card">
+                            <img class="large-front-thumbnail"
+                            src="{{ url_for('static', filename='imagenet_transform/'+unique_id+'.JPEG') }} "/>
+                        </div>
+                    </div>
+                </table>
+                <a class="btn btn-dark mb-2" href="/transformations" role="button">Back</a>
+            </div>
+        </div>
+    </div>
+{% endblock %}
+

app/templates/transformation_select.html

@@ -0,0 +1,36 @@
+{% extends "base.html" %}
+
+{% block content %}
+
+    <h1>Select Transformation and Image</h1>
+    <form action="" method="post" novalidate>
+        {{ form.hidden_tag() }}
+        <h4>
+            Transformation:
+        </h4>
+        <p>
+            Color:
+            {{ form.color() }}
+        </p>
+         <p>
+            Brightness:
+            {{ form.brightness() }}
+        </p>
+         <p>
+            Contrast:
+            {{ form.contrast() }}
+        </p>
+         <p>
+            Sharpness:
+            {{ form.sharpness() }}
+        </p>
+
+        <h4>
+            Image:
+        </h4>
+        <p>
+            {{ form.image }}
+        </p>
+        <button type="submit" class="btn btn-dark mb-2">Submit</button>
+    </form>
+{% endblock %}

config.py

@@ -9,6 +9,7 @@ class Configuration:
     # classification
     image_folder_path = os.path.join(project_root, 'app/static/imagenet_subset')
     models = ('resnet18', 'alexnet', 'vgg16', 'inception_v3',)
+
     # web server
     SECRET_KEY = os.environ.get('SECRET_KEY') or '9cj328s61hsd8'
     # queue

ml/classification_utils.py

@@ -8,8 +8,15 @@
 import os
 import time
 import torch
+import numpy as np
+import matplotlib.pyplot as plt
+import cv2
+import glob
 from PIL import Image
+
 from torchvision import transforms
+from PIL import Image
+from PIL import ImageEnhance
 
 from config import Configuration
 
@@ -40,7 +47,7 @@ def get_model(model_id):
     if model_id in conf.models:
         try:
             module = importlib.import_module('torchvision.models')
-            return module.__getattribute__(model_id)(pretrained=True)
+            return module.__getattribute__(model_id)(weights='DEFAULT')
         except ImportError:
             logging.error("Model {} not found".format(model_id))
     else:
@@ -84,3 +91,53 @@ def classify_image(model_id, img_id):
     img.close()
     time.sleep(5)
     return output
+
+
+def transformation_image(image, path, color_factor=1.0, brightness_factor=1.0, contrast_factor=1.0,
+                         sharpness_factor=1.0):  # default values
+    """Modify an input image sequentially adding enhancements saving the results in a specific directory"""
+
+    im = Image.open(image)
+    col = ImageEnhance.Color(im)
+    im_col = col.enhance(color_factor)
+    brh = ImageEnhance.Brightness(im_col)
+    im_col_brh = brh.enhance(brightness_factor)
+    con = ImageEnhance.Contrast(im_col_brh)
+    im_cal_brh_con = con.enhance(contrast_factor)
+    sharp = ImageEnhance.Sharpness(im_cal_brh_con)
+    im_cal_brh_con_sharp = sharp.enhance(sharpness_factor)
+    im_cal_brh_con_sharp.save(path)
+
+def plot_histogram(path, hist_path):
+    """A function to plot and save the histogram f a specific image and
+    save the histogram in a specific directory"""
+    im = cv2.imread(path)
+    vals = im.mean(axis=2).flatten()
+    counts, bins = np.histogram(vals, range(257))
+    plt.clf()   # to not overlap the images
+    plt.bar(bins[:-1] - 0.5, counts, width=1, edgecolor='none')
+    plt.xlim([-0.5, 255.5])
+    # plt.show()
+    saved_image = plt.savefig(hist_path)
+    return saved_image
+
+
+def plot_histogram_rgb(path, hist_path):
+    im = cv2.imread(path)
+    color = ('b', 'g', 'r')
+    for i, col in enumerate(color):
+        histr = cv2.calcHist([im], [i], None, [256], [0, 256])
+        plt.plot(histr, color=col)
+        plt.xlim([0, 256])
+    # plt.show()
+    saved_image = plt.savefig(hist_path)
+    plt.clf()  # to not overlap the images
+    return saved_image
+
+
+def clear_dir():
+    """A function to delete the content of a directory"""
+    files = glob.glob('app/static/imagenet_histogram')
+    for f in files:
+        os.remove(f)
+

prepare_models.py

@@ -12,7 +12,7 @@ def prepare_models():
         try:
             module = importlib.import_module('torchvision.models')
             # download model
-            _ = module.__getattribute__(model_name)(pretrained=True)
+            _ = module.__getattribute__(model_name)(weights='DEFAULT')
             del _  # free up memory
         except ImportError:
             logging.error("Model {} not found".format(model_name))