Photo Classification
One of the most important processes within CloudCard is Photo Classification. It occurs through the work of machine learning algorithms, known affectionately as Helper Bot. Photo Classification is the foundation for many of CloudCard's functions. Let's dive in and see what it's all about!
Confidence Score
Every photo that is submitted to CloudCard receives a confidence score from Helper Bot. The confidence score represents how certain Helper Bot is that the submitted photo will be approved. A lower score means that he is less uncertain of photo approval, while a higher score means he is more certain of photo approval.
Helper Bot has been trained with thousands of photos to identify a quality ID photo. The training taught him to spot different attributes within photos that would indicate approval or denial. These attributes are derived from standard ID photo requirements like shown in the screenshot below.
Helper Bot compares the photo that is submitted to the thousands of photos that he has been trained on. Based on how that photo stacks up to the other photos in his training, Helper Bot gives the photo a Confidence Score.
Helper Bot has given this photo a 65% confidence score because the cardholder submitted a photo that cropped off part of her head. He is unsure as to whether this photo should be approved or denied.
Status Thresholds
CloudCard uses Status Thresholds to categorize the photos that are are submitted. The photos are split into three different default categories based on the Confidence Score. There are default thresholds which I will show now. Be aware that these thresholds can be adjusted.
1. Approval threshold - set at an 90% or greater confidence score (green status bar).
2. Unsure threshold - set between a 20% and 90% confidence score (yellow status bar).
3. Denial threshold - set at a 20% confidence score or lower (red status bar).
Customized Status Thresholds
It's surprising how much ID photo standards differ from organization to organization. Some allow for a busy background and others want strictly white. The array of requirements is vast. So how does CloudCard accommodate for the differing photo standards of organizations?
We allow organizations to tune their status thresholds to fit their needs. For example, if an organization is more lenient with photo standards because they see their ID photo as a form of expression, we can make adjustments to the thresholds. The changes might look something like below:
- Approval Threshold - 65% or greater confidence score (green status bar).
- Unsure Threshold - between a 35% and 65% confidence score (yellow status bar).
- Denial Threshold - 35% or lower confidence score (red status bar)
Conversely, if an organization is a little more rigid with their photo requirements, we can tune the thresholds accordingly. These thresholds become very important when thinking about Auto Approval for your organization.