HS.Vision is:

Extract meaningful text descriptions of your images/videos, by exploiting the latest from the deep learning research.


JSON
[
  {
    "label": "sneakers",
    "score": 0.40725806451612906
  },
  {
    "label": "jeans",
    "score": 0.1774193548387097
  },
  {
    "label": "running shoes",
    "score": 0.14919354838709678
  }
]

Individuate and highlight in your image/video specific objects or patterns of interest.

JSON
{
  "labels": [ 
    {"label": "jacket/blazer", "score": 0.9951637}, 
    {"label": "pants/jeans", "score": 0.99527305},
    {"label": "shirt/blouse","score": 0.9879214},
    {"label": "skin", "score": 0.9800294}
  ],
  "original_size": [700, 1050],
  "bounding_box": [
    {"max_x": 609, "min_x": 302, "min_y": 164, "max_y": 535},
    {"max_x": 551, "min_x": 325, "min_y": 442, "max_y": 966},
    {"max_x": 476, "min_x": 397, "min_y": 162, "max_y": 449}
  ]
}

Understand the main colors in your image, or individuate precisely which colors are in your image, in order to be sure that the images mirror the reality with maximum fidelity.


JSON
"dominant_colors": [
  {
    "color": "Blue",
    "RGB": [55, 95, 194],
    "perc": 22
  },
  {
    "color": "Pink",
    "RGB": [248, 85, 160],
    "perc": 21
  },
  {
    "color": "Orange",
    "RGB": [252, 91, 5],
    "perc": 18
  },
  {
    "color": "Purple",
    "RGB": [144,96,197],
    "perc": 15
  },
  {
    "color": "Red",
    "RGB": [228,15,41],
    "perc": 14
  }
]

Extract textual content from your images/videos.


JSON
"pages_text_compact": [
  {
    "page": 0,
    "page_blurred": false,
    "data": [
	  "FAREWAY STORES, INC.",
	  "NORTH\t #061",
	  "REG. BAKED LAYS\t £\t 3.41",
	  "FASTCO PEAS 3202\t F\t 2.29",
	  "FASTCO PEAS 320Z\t F\t 2.29",
	  "40 $1.39 ea.",
	  "BDGT XWD NOODLES\t F\t 5.56",
	  "KR AMER WRAP 24\t F\t 1.88",
	  "KR AMER WRAP 24\t F\t 1.88",
	  "F FREE MUSH SOUP\t F\t 2.29",
	  "F FREE MUSH soup\t F\t 2.29",
	  "F FREE MUSH soup\t F\t 2.29",
	  "CAMP L-SLT MSHRM\t F\t 1.49",
	  "4@ $2.49 ea",
	  "STARKIST TUNA\t F\t 9.9",
	  "SUBTOTAL\t 35.63",
	  "TOTAL\t 35.63",
	  "VISA\t 35.63",
	  "ITEM COUNT 17",
	]
  }
]

The HS.Vision algorithms are lightweight, and can be mounted on edge computing devices (i.e., jetson cards). This allow to scale up a single computation to hundreds to thousands, on different devices, as it happens for a surveillance system with multiple cameras.

The image below shows the detection results on crowds of people, obtained by multiple cameras (boxes, points are related to different cameras), each one operating on a different edge device. HS.Vision is becoming Industry 4.0


Deep Learning

Under the hood, deep learning is made by neural networks which extract multi-level digital representations of the data, and evaluate them simultaneously, optimizing the best solution for the task at hand

Production-ready

Our solutions are available both in the cloud and on-premise, to ensure maximum security and scalability. Upload data, verify models, integrate results into your products or processes.

Professional service

Our team gives support and training in the use of the Humatics products, especially when it comes to ad-hoc solutions with custom-made extensions.