The Studio is perfect for analysing any still images and assets for any use case across any channel.
However, there are a few key points to consider when choosing which images to upload to the Studio to analyse. You must remember that the Studio and algorithm doesn't know what it is looking at - it simply looks at the composition of pixels in the image and predicts the attention in the scene based on a number of visual cues.
This means the Studio takes in all aspects of the image when predicting attention, including any distractors or visual noise from the image you may not want to be included.
Therefore it is important, that the images you upload are as true to real-life as possible.
Key Considerations
Some key points to consider generally when uploading assets to the Studio.
Resolution - Human eyes don't see pixelated images, higher the better.
Note: The Studio will automatically resize very high resolution images to a maximum of 1024x1024 pixels.
Context - Try to ensure your chosen images are as close to how your target audience will view your content, as Dragonfly AI replicates human vision.
Distractors - Exclude any parts of the image you don't want to be taken into account by the tool e.g. lighting, lens flares, borders etc.
Centre Bias - Does the image represent the final layout of elements in the scene? If not, and you want all assets to have the same chance of receiving attention regardless of location in the image, then make sure to turn off centre bias. Remember this is automatically turned on for every image uploaded on to the platform.
A/B/n testing - Ensure as much consistency as possible to allow for a fair test e.g. the same type of image at a similar resolution. If one image has a lot of background distractors and the other is just the asset, the results will be skewed, for example.
Some Key Examples
Real life imagery - eg a photo taken of your products on-shelf
The Studio replicates attention in a scene as if it were real human vision. So when taking your own photos and uploading them, to get a true understanding of how your products will perform in that context, it is vital for the image you use to be as close to natural human vision as possible. So what does this mean?
When taking the image in a store, you must think about what the average customer will see in that environment, and replicate this with your picture:
Landscape mode
Taken at eye-level
In focus
High resolution
Natural shot (no alterations e.g. fisheye lenses etc.)
Centre Bias on/off (is your product in its final/true position on-shelf?)
Lighting (is the lighting reflective of the average viewer? Make sure your flash is off)
This guide is true for any images you may be taking of your content in the real world, such as products in a store, billboards, point-of-sale material, and so on.
Product Images / Pack Shots
The Studio is replicating human vision, so this assumes your product will be taking up all/most of the audiences vision.
When uploading your pack shot, try and get the design as close to its final form as possible.
High resolution
All pixels in the scene will be analysed which will affect the attention distribution. Ensure all elements such as borders and watermarks are removed.
Take into account digital designs vs an image of the real thing. It is unlikely the real thing will look exactly the same as a digital design.
Any questions or need any further help? Please contact [email protected] or any member of the Dragonfly AI team.
The Customer team are happy to help check your images and run over any best practices with you. To set up a quick call contact your Customer Success manager or [email protected]