Skip to content

Latest commit

 

History

History
77 lines (52 loc) · 2.84 KB

object-detection-image-recon.md

File metadata and controls

77 lines (52 loc) · 2.84 KB
Raw

singnetlogo

Object Detection and Image Recognition

This service uses 2 other services:

  • yolov3-object-detection [Reference] to detect objects on images and;
  • cntk-image-recon [Reference] to classify these objects.

It is part of our third party DNN Model Services.

Welcome

This service is called MetaService because it calls other services to deliver a composed response.

What’s the point?

First the service calls yolov3-object-detection service and gets all detected objects from an image. Then it calls cntk-image-recon service for each object and returns its classification.

How does it work?

The user must provide the following inputs in order to start the service and get a response:

Inputs:

  • model_detect: DNN Model ("yolov3").
  • model_recon: DNN Model ("ResNet152").
  • img_path: An image URL.
  • confidence: Confidence of object detection (between 0 and 1).

You can use this service from SingularityNET DApp, clicking on SNET/MetaService.

You can also call the service from SingularityNET CLI (snet).

Assuming that you have an open channel to this service:

$ snet client call snet object-detection-image-recon default_group detect_recon '{"model_detect": "yolov3", "model_recon": "ResNet152", "img_path": "https://figopetinsurance.com/sites/default/files/styles/blog_detail/public/imagedogsman-and-dog-hiking-mountainsblog.jpg", "confidence": "0.5"}'
...
Read call params from cmdline...

Calling service...

    response:
        boxes: '[[340.5, 113.5, 153, 361], [432.5, 162.5, 73, 119], [256.5, 219.0, 133, 278]]'
        class_ids: '[0, 24, 16]'
        confidences: '[0.9899819493293762, 0.971903383731842, 0.9540891647338867]'
        delta_time: '202.7593548297882'
        top_1_list: '[''78.86%: Australian_cattle_dog'']'

What to expect from this service?

Input Image:

BackpackManDog Splash 1

with:

  • model_detect: yolov3
  • model_recon: ResNet152
  • confidence: 0.5

Response:

boxes           [[292.0, 143.0, 292, 278], [42.5, 60.0, 251, 346], [13.0, 129.5, 182, 203]]
confidences     [0.9982979893684387, 0.9968200922012329, 0.8117520809173584]
class_ids       [16, 0, 24]
top_1_list      ['80.91%: Australian_shepherd']

The service first got 3 bounding boxes (with a dog [16], a person [0] and a backpack [24]) using yolov3-object-detection.

Then it got the breed of the found dog (Australian_shepherd) using cntk-image-recon.