Technology

Top 3 automatically detected image errors

Automatic quality checks, also known as First Gate Checks, are carried out in real time each time a photo is uploaded. The Photo Collect AI analyzes the image for the most common errors. If the image is not OK, it will be rejected stating the error and an improved image will be requested. Our facts and figures show the range of rejection rates, how many images pass without any complaints on the second attempt and what the reasons for rejection are.

First, let's look at the key data: The photo of 7.5% of people is automatically rejected. Three-quarters of these people then submit an accepted photo on their second attempt, 17% need a third and 9% a fourth attempt.

On the fourth attempt, the AI is skipped and the image is accepted in any case. In this way, it can be excluded that physically impaired persons are denied the upload by the AI. If the photo quality is insufficient, the photo can still be rejected during manual quality control.

The total number of successful photo uploads, on the other hand, is only insignificantly reduced by the First Gate Checks: 99.4% of all people manage to submit a photo accepted by the AI.

Reasons for rejection

Rank 1: With almost 42%, the edge distance is the most frequent reason for rejection. This is classified as insufficient for a total of 3% of uploads. If the distance to the edge is too small, the image cannot be cropped nicely and gaps appear, usually below the chin or due to cut-off hairstyles. This problem often occurs with existing passport photos. Photo Collect defines a "Safety Zone" around the face, which must be filled with image content in any case.

The red rectangle (1, "Safety zone") must be filled with image content, this is missing at the top (2) and bottom (3).

Rank 2: A low-resolution face is criticized in 21% of rejections or 1.6% of all uploads. Photo Collect does not only look at the resolution of the submitted photo, but also at the number of pixels in the relevant face area. According to the ICAO standard, there must be a horizontal distance of at least 100 pixels between the pupils. This error occurs if the images are too low resolution from the outset (which is often the case with old image files), or if the person in the image is taken from too far away.

According to the ICAO standard, there must be at least 100 pixels between the pupils.

Rank 3: An incorrect head position is the trigger for 19% of rejections and occurs in 1.4% of all uploads. Here, Photo Collect distinguishes between the frequently occurring looking away to the left/right (“yaw”, 85%) and the shot from the frog's/bird's eye view (“pitch”, 15%). The rotation of the head is automatically corrected by Photo Collect and therefore does not appear in this statistic.

Looking too far to the right leads to undesirable results and is automatically rejected.

Interesting: Pictures with hygiene masks were widespread in 2021 - from mid-2022 we hardly find any such uploads. Group pictures with more than one person are extremely rare, they only occur in 0.09% of all uploads.

Conclusion

The quality checks have proven to be an effective and cost-effective means of filtering out obviously bad photos and avoiding later rejection in the manual quality control. Since the photo of 7.5% of people is automatically rejected at least once, we assume that the quality checks can reduce the rejection rate in the manual quality control by about one third (5 out of 15 percentage points).

Data used

The data originates from various Photo Collect instances from spring 2023. The analysis is based on approximately 23,000 photos.

Overall UploadsFrequencyReason
3.10%41.59%Margin
1.58%21.26%Resolution (Face)
1.41%18.98%Head Position
0.70%9.40%Blurred
0.43%5.72%Overexposed
0.09%1.17%Group
0.05%0.70%Image Noise
0.03%0.47%No Face
0.03%0.35%Image Quality
0.03%0.35%Underexposed

More blog posts