Updated vendors and added features

cloudy_vision.py

@@ -90,6 +90,44 @@ def render_from_template(directory, template_name, **kwargs):
     return template.render(**kwargs)
 
 
+def stringify_result(r):
+    s = r[0]
+    if r[1]:
+        s += "(%.2f)" % r[1]
+    if len(r) > 2 and r[2]:
+        s += "(" + r[2] + ")"
+    return s
+
+def render_tsv_report(image_results, vendor_stats, report_date):
+    # Building the report header (first row) based on the info from the
+    # first image result
+    header = "File" + "\t"
+    if 'image_tags' in image_results[0] and len(image_results[0]['image_tags']) > 0:
+        header += "Image tags" + "\t"
+    for vendor in image_results[0]['vendors']:
+        for feature_name, feature_results in vendor['standardized_result'].iteritems():
+            header += vendor['vendor_name'] + "_" + feature_name + "\t"
+    header += "\n"
+
+    ## Building the report body
+    report = ""
+    for image_result in image_results:
+        report += image_result['output_image_filepath'] + "\t"
+        if 'image_tags' in image_results[0] and len(image_results[0]['image_tags']) > 0:
+            # because the column was included or not based on the first image, we
+            # need to do this check first in order to keep the report columns aligned
+            if 'image_tags' in image_result and len(image_result['image_tags']) > 0:
+                report += ', '.join(image_result['image_tags']) + "\t"
+            else:
+                report += "\t"
+        for vendor in image_result['vendors']:
+            for feature_name, feature_results in vendor['standardized_result'].iteritems():
+                report += ', '.join(map(lambda r: stringify_result(r), feature_results)) + "\t"
+        report += "\n"
+
+    return header + report
+
+
 def vendor_statistics(image_results):
     vendor_stats = {}
 
@@ -259,6 +297,17 @@ def process_all_images():
     with open(output_html_filepath, 'w') as output_html_file:
         output_html_file.write(output_html.encode('utf-8'))
 
+    # Create TSV report with all results
+    output_tsv = render_tsv_report(
+        image_results,
+        vendor_stats,
+        datetime.datetime.today()
+    )
+
+    # Write TSV output
+    output_tsv_filepath = os.path.join(settings('output_dir'), 'output.tsv')
+    with open(output_tsv_filepath, 'w') as output_tsv_file:
+        output_tsv_file.write(output_tsv.encode('utf-8'))
 
 if __name__ == "__main__":
     process_all_images()

example_api_keys.json

@@ -1,7 +1,15 @@
 {
     "google" : "YOUR KEY HERE",
-    "microsoft" : "YOUR KEY HERE",
-    "ibm" : "YOUR KEY HERE",
+    "microsoft" : {
+      "location": "YOUR LOCATION HERE (eg. canadacentral, westus2...)",
+      "api_key": "YOUR KEY HERE",
+    },
+    "ibm" : {
+        "api_key": "YOUR KEY HERE",
+        "url": "https://gateway.watsonplatform.net/visual-recognition/api",
+        "language": "en"
+      },
+    "clarifai": "YOUR KEY HERE",
     "cloudsight" : {
         "api_key" : "YOUR KEY HERE",
         "api_secret" : "YOUR SECRET HERE"

readme.md

@@ -15,7 +15,7 @@ Run a corpus of images through multiple computer vision API vendors. View image
 1. Optionally match the tags returned with your desired tags to test accuracy.
 1. Calculate stats around response times, number of tags returned, etc.
 1. Create a scaled copy of the original image with height 200px.
-1. Generate output.html to show all the images and labeling results in an easy to consume manner.
+1. Generate output.html and output.tsv to show all the images and labeling results in an easy to consume manner.
 
 ## Usage
 
@@ -30,9 +30,6 @@ Run a corpus of images through multiple computer vision API vendors. View image
 
 The keys should not be placed in the api_keys.json file but in ~/.aws/credentials and ~/.aws/config. See http://docs.aws.amazon.com/cli/latest/userguide/cli-chap-getting-started.html#cli-config-files
 
-### Note for Clarifai
-
-The keys should not be placed in the api_keys.json file but in ~/.clarifai/config. See https://github.com/Clarifai/clarifai-python#setup
 
 ## Desired Tags
 

requirements.txt

@@ -1,7 +1,7 @@
 numpy==1.12.1
 cloudsight==0.9.2
-clarifai==2.0.21
-watson-developer-cloud==0.25.2
+clarifai==2.2.3
+watson-developer-cloud==1.6.0
 Jinja2==2.9.6
 boto3==1.4.4
-Pillow==2.9.0
+Pillow==2.9.0

static/template.html

@@ -84,10 +84,24 @@ <h2 class="image_name">desired_tags:
               <td class="result_values">
                   {% for result in feature_results %}
                   {% set score = result[1] %}
+                  {% set annotation = result[2] %}
+
+                  {% if feature_name == 'colors' %}
+                  <span class="result">
+                  {% else %}
                   <span class="result" style="opacity: {{ score }}">
-                      {{ result[0] }}
+                  {% endif %}
+                      {% if feature_name == 'colors' %}
+                        <span style="background-color: {{ result[0] }};">{{ result[0] }}</span>
+                      {% else %}
+                        {{ result[0] }}
+                      {% endif %}
+
                       {% if score is not none %}
-                      <span class="score">({{ ("%.2f" % score).lstrip('0') }})</span>
+                        <span class="score">({{ ("%.2f" % score).lstrip('0') }})</span>
+                      {% endif %}
+                      {% if annotation %}
+                        <span>({{ annotation }})</span>
                       {% endif %}
                       {% if not loop.last %}
                       ,
@@ -128,7 +142,7 @@ <h2 class="image_name">desired_tags:
     {% endfor %}
 
     {% if vendor_stats|length > 0 %}
-    <div class="row" id="statistics">    
+    <div class="row" id="statistics">
         <div class="twelve columns result">
             <h4>Stats</h4>
             <p>

vendors/clarifai_.py

@@ -3,7 +3,7 @@
 
 
 def call_vision_api(image_filename, api_keys):
-    app = ClarifaiApp()
+    app = ClarifaiApp(api_key=api_keys['clarifai'])
     model = app.models.get('general-v1.3')
     image = ClImage(file_obj=open(image_filename, 'rb'))
     result = model.predict([image])

vendors/google.py

@@ -41,6 +41,10 @@ def call_vision_api(image_filename, api_keys):
               "type": "SAFE_SEARCH_DETECTION",
               "maxResults": 10
             },
+            {
+              "type": "IMAGE_PROPERTIES",
+              "maxResults": 10
+            },
           ]
         }
       ]
@@ -71,4 +75,10 @@ def get_standardized_result(api_result):
         for annotation in api_result['logoAnnotations']:
             output['logo_tags'].append((annotation['description'], annotation['score']))
 
+    if 'imagePropertiesAnnotation' in api_result and 'dominantColors' in api_result['imagePropertiesAnnotation']:
+        output['colors'] = []
+        for color_annotation in api_result['imagePropertiesAnnotation']['dominantColors']['colors']:
+            color = "#{0:x}".format(int(color_annotation['color'].get('red') or 0)) + "{0:x}".format(int(color_annotation['color'].get('green') or 0)) + "{0:x}".format(int(color_annotation['color'].get('blue') or 0))
+            output['colors'].append((color, color_annotation['score']))
+
     return output

vendors/ibm.py

@@ -1,19 +1,26 @@
 from watson_developer_cloud import VisualRecognitionV3
 
 def call_vision_api(image_filename, api_keys):
-    api_key = api_keys['ibm']
+    api_key = api_keys['ibm']['api_key']
+    iam_url = api_keys['ibm']['url']
+    language = api_keys['ibm']['language']
 
-    # Via example found here:
-    # https://github.com/watson-developer-cloud/python-sdk/blob/master/examples/visual_recognition_v3.py
-    visual_recognition = VisualRecognitionV3('2016-05-20', api_key=api_key)
+    # Via example found in the code snippets from the Implementation doc
+    # in the IBM Watson dashboard/console
+    visual_recognition = VisualRecognitionV3('2018-03-19',
+        iam_api_key=api_key,
+        url=iam_url)
 
     with open(image_filename, 'rb') as image_file:
-        result = visual_recognition.classify(images_file=image_file)
+        result = visual_recognition.classify(images_file=image_file, headers={'Accept-Language': language})
 
     return result
 
 
 def get_standardized_result(api_result):
+    color_indicators = ['color', 'couleur']
+        # add color word in other languages if you use a different
+        # language, as per the configs in api_keys
     output = {
         'tags' : [],
     }
@@ -28,4 +35,18 @@ def get_standardized_result(api_result):
         for tag_data in api_result['classes']:
             output['tags'].append((tag_data['class'], tag_data['score']))
 
+        # some tags contain the word "color", so we use that indicator
+        # to extract color information. This is not as reliable as other
+        # vendors providing structured information about colors, but it
+        # is better than nothing
+        color_tags = [tag for tag in api_result['classes'] for color_indicator in color_indicators if color_indicator in tag['class']]
+        if len(color_tags) > 0:
+            output['colors'] = []
+            for color_tag in color_tags:
+                color = color_tag['class']
+                for color_indicator in color_indicators:
+                    color = color.replace(color_indicator, '')
+                color = color.strip()
+                output['colors'].append((color, color_tag['score']))
+
     return output

vendors/microsoft.py

@@ -2,8 +2,9 @@
 import requests
 
 def call_vision_api(image_filename, api_keys):
-    api_key = api_keys['microsoft']
-    post_url = "https://api.projectoxford.ai/vision/v1.0/analyze?visualFeatures=Categories,Tags,Description,Faces,ImageType,Color,Adult&subscription-key=" + api_key
+    api_key = api_keys['microsoft']['api_key']
+    location = api_keys['microsoft']['location']
+    post_url = "https://" + location + ".api.cognitive.microsoft.com/vision/v1.0/analyze?visualFeatures=Categories,Tags,Description,Faces,ImageType,Color,Adult&subscription-key=" + api_key
 
     image_data = open(image_filename, 'rb').read()
     result = requests.post(post_url, data=image_data, headers={'Content-Type': 'application/octet-stream'})
@@ -30,25 +31,37 @@ def get_standardized_result(api_result):
     output = {
         'tags' : [],
         'captions' : [],
-#        'categories' : [],
+        'categories' : []
 #        'adult' : [],
 #        'image_types' : []
 #        'tags_without_score' : {}
     }
 
-    for tag_data in api_result['tags']:
-        output['tags'].append((tag_data['name'], tag_data['confidence']))
+    if 'description' in api_result:
+        if 'tags' in api_result['description']:
+            for tag_data in api_result['description']['tags']:
+                output['tags'].append((tag_data, None))
+                    # tags do not have confidence or scores
 
-    for caption in api_result['description']['captions']:
-        output['captions'].append((caption['text'], caption['confidence']))
+        if 'captions' in api_result['description']:
+            for caption in api_result['description']['captions']:
+                output['captions'].append((caption['text'], caption['confidence']))
 
-#    for category in api_result['categories']:
-#        output['categories'].append(([category['name'], category['score']))
+    if 'categories' in api_result:
+        for category in api_result['categories']:
+            output['categories'].append((category['name'], category['score']))
 
-#    output['adult'] = api_result['adult']
+    if 'color' in api_result:
+        output['colors'] = []
+        for color_type in ['accentColor', 'dominantColorBackground', 'dominantColorForeground']:
+            if color_type in api_result['color']:
+                output['colors'].append((api_result['color'][color_type], None, color_type))
+                    # colors do not have confidence or scores
+        if 'dominantColors' in api_result['color']:
+            for color in api_result['color']['dominantColors']:
+                output['colors'].append((color, None, 'dominant'))
 
-#    for tag in api_result['description']['tags']:
-#        output['tags_without_score'][tag] = 'n/a'
+#    output['adult'] = api_result['adult']
 
 #    output['image_types'] = api_result['imageType']