Automated image analysis classifies all member-uploaded images based on the elements detected within the image. This powerful automated tool categorizes elements into Object/Activity, Landmark, Brand/Logo, Text, and Face, then provides additional descriptions of those elements detected in the photo. Furthermore, when image analysis detects a face, the emotion associated with it will also be analyzed.
The image data in the Image Analysis tab within an Activity Export helps you quickly summarize images at scale.
The following data will be available in the Image Analysis tab of your export spreadsheet:
- Image Analysis ID—Unique ID that is given to an image when it is run through automated analysis
- Uploaded File ID—Unique ID that is given to an image when a member uploads the file to Fuel Cycle
- File Name—Filename of the image that you can use to cross-reference data against the original file source.
- Content ID—Unique ID of the activity in which the file was uploaded.
- Content Title—Activity title in which the file was uploaded.
- Comment ID—Unique ID of the comment if the file was attached to a comment.
- Comment Message—If the file was attached to a comment, the comment's text will display in this column.
-
Type—All elements detected during the automated analysis will be categorized into the following types:
- Object/Activity
- Landmark
- Brand/Logo
- Text
- Face
- Safe Search—Safe search is run on all images to determine if there is anything inappropriate in the image. The results of this scan are shown in the columns Adult-Racy (see below for description)
- Description—For each element found on the page, a new row is included in the sheet and this column describes each element found, such as “Bike” if a bike is in the photo.
- Score—The confidence number at which the automated service believes the element to be what is described. The closer the score is to 1.0, the higher the likelihood that the item is represented in the photo. For example, if the score is 0.983136, the element is likely in the photo, but a score of .149603 would be very unlikely.
- Latitude/Longitude—When a landmark is found in the image, this column will give the latitude and longitude of the landmark.
-
Adult, Spoof, Medical, Violence, Racy—These columns are related to the Safe Search scan and will show a range from VERY_UNLIKELY to VERY_LIKELY to be deemed inappropriate based on those 5 categories.
Images deemed inappropriate will not be automatically removed from the community at this time. This feature will be added in the future, but for now, you will have to remove images manually if they are inappropriate.
- Anger, Headwear, Joy, Sorrow, Surprise: These columns are related to facial detection and show a range from VERY_UNLIKELY to VERY_LIKELY to be the emotion based on those 5 types.
- Text—If any text has been detected in an image, this column will include the text as it has been detected.
Use Cases
Image Analysis will help you manage activities where you ask members to upload a lot of images. Rather than having to manually view and tag each image, this gives you a shortcut to sorting the images based on the detected elements and will ultimately help you find insights and themes quickly.
Some of the common use cases include:
-
Themes
Quickly discover common objects, activities, and themes across all images uploaded to an activity without having to look at any images individually. -
Brand Mentions
Find out how often your brand or your competitor brands show up in the images attached to an activity. -
Emotions
Get an immediate sense of the overall emotion of the images in an activity. -
Readable/searchable Text
Text in an image is converted into readable text that you can then use in searching, sorting, and filtering.