OD & Counter Inference Service: COCO format with Metrics
An Object Detector and Counter API service, capable of detecting and counting Common Objects (80 Object classes described in the COCO Dataset), from the given Image file
Microservice with Metrics related to Artificial Intelligence in the context of Automotive Manufacturing Electronics Manufacturing & Equipment Food and Beverage Manufacturing
Provided by DFKI (German Research Center for Artificial Intelligence) 9 months ago (last modified 5 months ago); viewed 152 times and bound 1 time
Description:
The Object Detector & Counter API service, which could take any Image file as Input and gives a JSON file consisting of the name of the objects, bounding boxes, confidence scores and the number of instances, as response.
The user can send the Input image(s) to the API service using normal POST requests.
Users can communicate to the Microservice on the port 5050, with the `/detect/` Endpoint, using a POST request.
http://<Exposed-IP>:5050/docs will shows the available Endpoints on this service using Swagger UI (please fill the correct exposed IP address of your deployment instead of <Exposed-IP>)
The response shall be in the following format (JSON):
{
"filename": "...",
"detection results": {
"detected objects":{
"label": [...],
"confidence": [...],
"boxes": [[x,y,w,h], [x1,y1,w1,h1], ...] },
"counts":{
"class name": "no of instances", ... },
"image height": ...,
"image width": ... } }
This Microservice generates Two Metrics: probability and label, which are corresponding to the confidence score of the prediction and the name of the object/predicted class. These Metrics can be tracked by the users to produce necessary alerts.
_________________________________________________________________________________________________________________________________________________________________
!!! Important Notes !!!
- This Microservice is configured to communicate to Message Handler and to track these two Metrics by default. !!
- The microservice mounts the Model files from the /odapi/od_app/model folder and therefore the Model file converted in onnx format with the filename yolov5s.onnx and target class file with the name coco.names shall be available on this /odapi/od_app/model folder to make this microservice work. !!!
use the volume mapping in your custom deployments:
For Example:
- ./data:/odapi/od_app/model
Classification: Machine Learning
Type: Deep Learning
Available Metrics
label
Name of the objects detected
- Unit: No Unit
probability
The confidence score at which the AI application identifies and classifies the objects as a particular class
- Unit: No Unit
Container Deployment Information
- Format: docker-compose
- GUI Microservice: No
- Workload type: service
- Opened ports: