Car Picture Quality Control
The Car Picture Quality Control endpoint allows developers to verify the quality of images before sending them to ALTO AI for further analysis. This is an important step to ensure that the AI is able to accurately detect and analyze any defects or damages present in the images.
{ "imageQualityCompliance": "not_good", "reason": { "blur": 0.5586, "corrupted": 0, "dark": 0.007, "dirt": 0.0004, "flare": 0.0645, "reflection": 0.0382, "screen": 0.0564, "shadow": 0.0257, "snow": 0.0803, "wet": 0.0454 } }
Prompt user to upload a better picture for a reliable damage detection
The Car Picture Quality Control endpoint evaluates multiple criteria to ensure images meet the required quality standards before being processed for damage detection.
What This Endpoint Checks
Resolution & Size Compliance: Confirms that the image meets the minimum resolution and size requirements necessary for accurate AI analysis, ensuring reliable damage detection.
Visual Quality: Assesses brightness, blurriness, and other key visual factors.
Vehicle Verification: Ensures the image contains a car or utility vehicle.
Additional Details: Provides insights such as the car’s angle, whether the image is cropped, and whether it is a close-up or a distant shot.
💡 Good to know: This functionality is seamlessly integrated into our Shoot Inspect feature.
How it works
Two AI models work together to evaluate image quality:
1. Feature Detection Model
This model assigns confidence scores (ranging from 0 to 1) for the presence of specific visual elements, reported in the reason object.
The score does not indicate whether the feature makes the image unsuitable for damage detection, only that it is present.
Example: A 0.90 score for "wet" means the AI is 90% sure the car body is wet, but this alone does not necessarily make the image non-compliant for damage detection. For example it does not mean that 90% of the vehicle is wet.
2. Image Compliance Model
This model evaluates the overall compliance of the image with the required quality standards for reliable damage detection. It returns three possible values:
not_good – The image is not suitable for damage detection because damages may be obscured, hidden or invisible in the picture.
exploitable – Some factors may slightly interfere with detection, but the image is still good enough for the AI to identify most visible damages.
good – The image meets all quality requirements and is ideal for damage detection.
Additionally, an overall quality score is provided, where a higher score indicates better compliance with quality standards.
When to Reject Photos
📌 It is recommended to reject photos where imageQualityCompliance is set to "not_good". Before rejecting, review the reason object to understand which issues may affect AI analysis.
🚨 If the type field returns "not a car", it means no vehicle (or part of a vehicle) was detected, making the image unsuitable for further AI analysis.
Recommendations
To ensure faster results:
✅ Compress images to reduce file size.
✅ Resize images to 512 pixels.
Analyze the Quality of One Image - Endpoint
POST https://alto.tchek.fr/apiV1/quality/image
Request Body
url
String
example: { "url": "https://tchek-images.s3.eu-west-3.amazonaws.com/swagger/qi_damaged_vehicle.jpeg" }
Sample picture:
https://tchek-images.s3.eu-west-3.amazonaws.com/swagger/qi_damaged_vehicle.jpeg"
Response
{
"id": "697f5a68acf75f4d",
"status": 200,
"data": [
{
"classification": {
"imageQualityCompliance": "not_good",
"angle": "Side Driver",
"cut": "not cut",
"quality": 0.0678,
"reason": {
"dark": 0.46295,
"dirt": 0.539551,
"flare": 0.74524,
"obstacle": 0.97412,
"reflection": 0.889648,
"screen": 0,
"shadow": 0.69519,
"sticker": 0.319824,
"wet": 0.172607,
"blur": 0,
"compression": 0.906537,
"corrupted": 0,
"height": 1024,
"width": 576
},
"rotation": "0",
"type": "distant"
},
"status": 200,
"url": "image_URL"
}
]
}
data.classification
Name
Description
Example
imageQualityCompliance
The overall image quality compliance evaluation of the image issued by Alto AI
not_good, good, exploitable
quality
The overall image quality score issued by Alto AI
0 to 1. 0 = not good 1= good
angle
The shooting angle of the vehicle
"Front"
"Front t34 Driver"
"Side Driver"
"Back t34 Driver"
"Back"
"Back t34 Passenger"
"Side Passenger"
"Front t34 Passenger"
The ID matching is here: LINK
Please note that those angles refer to left-hand drive vehicles.
cut
Whether the entire side of the vehicle fits in the picture.
"not cut", "cut", or "border" when the car fits exactly within the picture frame
rotation
Whether the picture is rotated
"0": The picture is not rotated
"90_CLOCKWISE": The picture is rotated 90° clockwise
"ROTATE_180": The picture is rotated 180° clockwise
"90_COUNTERCLOCKWISE": The picture is rotated 90° counter-clockwise
type
Whether the vehicle in the picture is taken from afar or from up close.
"distant": picture is taken from afar
"close": picture is taken from up close
"interior": the interior of the vehicle is detected in the picture
"not_a_car": no car was detected in the picture
"cargo_area": whether the cargo area of an utility vehicle was detected in the picture.
Reasons
Attribute
Description
Possible values
blur
Indicates the confidence level of Alto AI in detecting blurriness in the image. A high score suggests a strong presence of blur, but does not necessarily mean the image is unusable.
0 to 1.
0 = ALTO AI has no confidence that the feature "blur" is present in the image. However, this does not necessarily mean the image is sharp.
1 = ALTO AI is fully confident that the feature "blur" is present in the image. However, this does not necessarily mean the image is unusable or of poor quality.
compression
Indicates the confidence level of Alto AI in detecting compression artifacts in the image. A high score means noticeable compression artifacts are present.
0 to 1.
0 = ALTO AI has no confidence that the feature "compression" is present in the image. However, this does not necessarily mean the image is compressed.
1 = ALTO AI is fully confident that the feature "compression" is present in the image. However, this does not necessarily mean the image is too compressed or of poor quality.
corrupted
Determines whether the image file is corrupted. A score of 1 indicates a corrupted image, while 0 means the image is intact.
0 to 1.
0 = image not corrupted
1= image is corrupted
dark
The confidence score issued by Alto AI concerning darkness being present in the picture
0 to 1.
0 = ALTO AI has no confidence that the feature "dark" is present in the image. However, this does not necessarily mean the image is well lit.
1 = ALTO AI is fully confident that the feature "dark" is present in the image. However, this does not necessarily mean the image is unusable or too dark.
dirt
The confidence score issued by Alto AI concerning dirt being present in the picture
0 to 1.
0 = ALTO AI has no confidence that the feature "dirt" is present in the image. However, this does not necessarily mean the car is not dirty.
1 = ALTO AI is fully confident that the feature "dirt" is present in the image. However, this does not necessarily mean the image is unusable or the car too dirty.
flare
The confidence score issued by Alto AI concerning flares caused by strong light sources being present in the picture
0 to 1.
0 = ALTO AI has no confidence that the feature "flare" is present in the image.
1 = ALTO AI is fully confident that the feature "flare" is present in the image. However, this does not necessarily mean the image is unusable.
reflection
The confidence score issued by Alto AI concerning reflections being present in the picture, which could obscure details and impact analysis.
0 to 1.
0 = ALTO AI has no confidence that the feature "reflection" is present in the image.
1 = ALTO AI is fully confident that the feature "reflection" is present in the image. However, this does not necessarily mean the image is unusable.
screen
The confidence score issued by Alto AI concerning the probability that the image in the picture is the photograph of a screen (e.g., a monitor or smartphone display).
0 to 1.
0 = original picture
1= picture taken off of a screen
shadow
The confidence score issued by Alto AI concerning shadows being present in the picture, which may impact visibility and damage detection.
0 to 1.
0 = ALTO AI has no confidence that the feature "shadow" is present in the image. However, this does not necessarily mean the image is well exposed.
1 = ALTO AI is fully confident that the feature "shadow" is present in the image. However, this does not necessarily mean the image is unusable or of poor quality.
sticker
The confidence score issued by Alto AI concerning stickers being present in the picture which may cover parts of the photographed object
0 to 1.
0 = ALTO AI has no confidence that the feature "sticker" is present in the image.
1 = ALTO AI is fully confident that the feature "sticker" is present in the image. However, this does not necessarily mean the image is unusable.
wet
The confidence score issued by Alto AI concerning wet surfaces in the image, potentially affecting damage visibility.
0 to 1.
0 = ALTO AI has no confidence that the feature "wet" is present in the image.
1 = ALTO AI is fully confident that the feature "wet" is present in the image. However, this does not necessarily mean the image is unusable or the car is too wet to be analysed.
snow
The confidence score issued by Alto AI concerning snow flakes being present in the picture, potentially affecting damage visibility.
0 to 1.
0 = ALTO AI has no confidence that the feature "snow" is present in the image.
1 = ALTO AI is fully confident that the feature "snow" is present in the image. However, this does not necessarily mean the image is unusable or the car is too wet to be analysed.
obstacle
The confidence score issued by Alto AI concerning obstacles between the camera and the vehicle, potentially affecting damage visibility.
0 to 1.
0 = ALTO AI has no confidence that the feature "obstacle" is present in the image.
1 = ALTO AI is fully confident that the feature "obstacle" is present in the image. However, this does not necessarily mean the image is unusable or the car is not visible enough to be analysed.
height
The height resolution of the image, in pixels
integer
width
The width resolution of the image, in pixels
integer
Analyze the Quality of a List of Images - Endpoint
POST https://alto.tchek.fr/apiV1/quality/image/list
Request Body
urls
Array
List of file URLs . Ex: { "urls": [ "https://tchek-images.s3.eu-west-3.amazonaws.com/swagger/damagedVehicleT34DriverSide.jpeg", "https://tchek-images.s3.eu-west-3.amazonaws.com/swagger/qi_damaged_vehicle.jpeg" ] }
{
"id": "bee89a1390da8cd4",
"status": 200,
"data": [
{
"classification": {
"angle": "Front t34 Driver",
"cut": "not cut",
"imageQualityCompliance": "good",
"quality": 0.8223,
"reason": {
"Dark": 0.2261,
"Dirt": 0.7715,
"Flare": 0.1724,
"Reflection": 0.0454,
"Shadow": 0.2181,
"Sticker": 0.1393,
"Wet": 0.1865,
"blur": 0,
"compression": 0.9064,
"corrupted": 0,
"height": 3024,
"obstacle": 0.097,
"screen": 0.0099,
"width": 3072
},
"rotation": "0",
"type": "distant"
},
"status": 200,
"url": "https://tchek-images.s3.eu-west-3.amazonaws.com/swagger/damagedVehicleT34DriverSide.jpeg"
},
{
"classification": {
"angle": "Front t34 Driver",
"cut": "cut",
"imageQualityCompliance": "not_good",
"quality": 0.161,
"reason": {
"Dark": 0.8633,
"Dirt": 0,
"Flare": 0.0962,
"Reflection": 0.8237,
"Shadow": 0.0873,
"Sticker": 0,
"Wet": 0,
"blur": 0.95,
"compression": 0.4071,
"corrupted": 0,
"height": 1152,
"obstacle": 0.0588,
"screen": 0.0564,
"width": 2048
},
"rotation": "0",
"type": "distant"
},
"status": 200,
"url": "https://tchek-images.s3.eu-west-3.amazonaws.com/swagger/qi_damaged_vehicle.jpeg"
}
]
}
Last updated
Was this helpful?