TODO: Module I/O

[bleggett]

Since this module wraps all image capture and detection - it should spit out a simple list of detected objects for consumption by other node-red modules, e.g.

[
  { key: "person", value: 2 }
  { key: "cellphone", value: 2 }
]

[jbondpk]

@bleggett - So goal is to produce a simple list for consumption by other future node-red modules. As background, here's an example response currently produced for three bananas.

[{
    "detections": [
        {
            "bbox": [
                38,
                608,
                529,
                1003
            ],
            "class": "banana",
            "score": 0.9932200908660889
        },
        {
            "bbox": [
                255,
                332,
                768,
                703
            ],
            "class": "banana",
            "score": 0.9972244501113892
        },
        {
            "bbox": [
                55,
                52,
                566,
                443
            ],
            "class": "banana",
            "score": 0.998857855796814
        }
    ],
    "image": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL/2wBDAQkJCQwLDBgNDRgyIRwhMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjL/wAARCAQAAwADASIAAhEBAxEB/8QAHwAAAQUBAQEBAQEAAAAAAAAAAAECAwQFBgcICQoL/8QAtRAAAgEDAwIEAwUFBAQAAAF9AQIDAAQRBRIhMUEGE1FhByJxFDKBkaEII0KxwRVS0fAkM2JyggkKFhcYGRolJicoKSo0NTY3ODk6Q0RFRkdISUpTVFVWV1hZWmNkZWZnaGlqc3R1dnd4eXqDhIWGh4iJipKTlJWWl5iZmqKjpKWmp6ipqrKztLW2t7i5usLDxMXGx8jJytLT1NXW19jZ2uHi4+Tl5ufo6erx8vP09fb3+Pn6/8QAHwEAAwEBAQEBAQEBAQAAAAAAAAECAwQFBgcICQoL/8QAtREAAgECBAQDBAcFBAQAAQJ3AAECAxEEBSExBhJBUQdhcRMiMoEIFEKRobHBCSMzUvAVYnLRChYkNOEl8RcYGRomJygpKjU2Nzg5OkNERUZHSElKU1RVVldYWVpjZGVmZ2hpanN0dXZ3eHl6goOEhYaHiImKkpOUlZaXmJmaoqOkpaanqKmqsrO0tba3uLm6wsPExcbHyMnK0tPU1dbX2Nna4uPk5ebn6Onq8vP09fb3+Pn6/9oADAMBAAIRAxEAPwDo5lAC5PV3X8xVNiwtCckFX29ff/69XrvAcLnGJgc/UVSJ3K6n+Lb39xXGdZsREeSh/wBjpiqYz/ZsyYyV3gfnmrI/1MYBzgGq+AYLvb6t/KmIHibGDjZskY/Tg1NcMz3EM2QMk9O1R35K6bFt4LIQf++TgUkxKx2bngNwR+Gf...",
    "things": [
        {
            "thing": "banana"
        },
        {
            "thing": "banana"
        },
        {
            "thing": "banana"
        }
    ]
}

Along with the original detections member produced with all the detail, had added the rendered image member with bounded boxes and the things member in particular for consumption of the Node-RED inventory flow. In particular the things member lists each thing instance separately so the inventory flow subtraction/difference node could perform this difference. It currently requires this form to do an "individual" difference of things/items.

We could add another member to the existing response to further summarize the objects as you've listed. Had shifted to the terms thing and count as a more user friendly terms. If we want to add this, would we want something like:

{
  "detections": ...,
  "image": ...,
  "things": ...,
  "thing-counts": {
    [ 
      { "thing": "person", "count": 2 },
      { "thing": "cellphone", "count": 2 }
    ]
  }
}

Thoughts?

[bleggett]

@jbondpk That makes sense - there can be the simple, top level summary fields, and then a details property at the top level that expands into the full result set you have there.