Detect multiple objects and locate them inside an image using a Detection Transformer
Upload a normal non-confidential image and click Detect Objects with AI.
Example 1:
Upload a street image containing cars and people.
Example 2:
Upload an image containing a person sitting on a chair.
Example 3:
Upload an image containing several objects such as a
laptop, keyboard, bottle and cup.
Example 4:
Upload a parking-area image containing several vehicles.
Example 5:
Upload a kitchen or office image containing multiple
everyday objects.
Example 6:
Upload a photo containing one large object and several
smaller objects and observe the different bounding boxes.
Example 7:
Upload a public image of an electrical connector and
observe whether the general-purpose model recognizes it.
Example 8:
Upload a public wiring-harness image and observe the
limitation of a model that was not specifically trained
for wiring-harness components.
Unlike image classification, object detection does not only try to understand the overall image.
The AI attempts to identify multiple individual objects and determine their positions inside the image.
Each detection normally contains three important pieces of information:
Class:
What object the AI thinks it detected.
Score:
How strongly the model supports that detection.
Bounding Box:
The object's estimated location in the image.
Image
↓
"This image contains a car."
Image
↓
"There are 3 cars and 2 people,
and these are their locations."
A specialized object-detection model could potentially be trained to recognize engineering component classes.
For example, a wiring-harness-specific training dataset could contain annotated images of connectors, clips, terminals, grommets, ground eyelets and other components.
The important difference is that the AI would need appropriate training data for those specialized classes.
This demonstration uses a general-purpose object detection model.
It is not trained specifically for automotive wiring-harness engineering.
Therefore, do not expect reliable detection of specialized components such as harness clips, terminals, ground eyelets, splices or connector families.
Use only public, personal, synthetic or otherwise non-confidential images in this public demo.