Smart annotation

Embedded Machine Learning and Computer Vision algorithms
to accelerate the annotation process

Try our demo below !

Try to annotate objects in this demo with a single user click.
Change the scale of the region of interest with + / – keys, and click on the objects you want to annotate.
The object detector is a TensorFlow.js implementation of SSD-MobileNet trained on MsCoco dataset with 82 classes.

Easily write your own annotation application in Javascript using the smart rectangle element:

Copy to Clipboard

import { css, html, LitElement} from 'lit-element';
import '@pixano/graphics-2d';

const colorMap = {
  'person': 'blue',
  'bicycle': 'red',
  'car': 'green',
  ...
};

class MySmartDemo extends LitElement {

onCreate(evt) {
    // listening to the create event dispatched
    // by the element to assign a nice color to
    // the new rectangle.
    const newObj = evt.detail;
    newObj.color = colorMap[newObj.category];
  }

get element() {
    // Utility getter of the element
    return this.shadowRoot.querySelector('pxn-smart-rectangle');
  }

render() {
    // Render the template with the rectangle element
    // enriched with some buttons to interact with it.
    return html`
    <pxn-smart-rectangle image="image.jpg" @create=${this.onCreate}></pxn-smart-rectangle>
    <div>
        <button @click=${() => this.element.viewControls.zoomIn()}>+</button>
        <button @click=${() => this.element.viewControls.zoomOut()}>-</button>
    </div>`;
  }
}

customElements.define('my-smart-demo', MySmartDemo);

Browse the complete code on GitHub

Smart annotation

Third party