Image dataset annotation according to your model's strict requirements


Product shelves dataset annotation
The potential of machine learning is unlimited. Unfortunately, a lack of training data can limit your application.

Data collection is a challenging task in and of itself. Even if it is already solved, image annotation is also complex and resource-intensive.

In the following case, we received a task to collect and annotate a grocery shelves image dataset.
The original annotation guideline was almost 20 pages long. We have reduced it to a few points to describe the job briefly.


Task Formulation

For annotating a container, cap, logo, and price tag objects, you should use a tight bounding:

All four sides of the box should touch the container's boundaries in the image.

The containers should be labeled with the product type.

The logo should be labeled with the brand name.

The price tag should be labeled with the price.

Each container - logo pair must be united by a common attribute.

For each container cap, the attribute should be defined. The attribute must indicate which SKU the cover belongs to.

All containers must be combined with a common attribute with a corresponding price tag.

FORMULATION

Free trial



We always begin our collaborations with a trial task. There are two reasons for this.

First, you can evaluate the quality of our work, and make your comments if required.

Second, we can determine if everything is clear to us, ask the questions that have arisen, approve the file format, and decide what is the best way to interact.

Usually, after the trial task, the customer supplements or slightly corrects the requirements.

FREE TRIAL

Completing the task

During the order's execution, edge cases may arise that we did not encounter during the trial. In each such case, the project manager contacts the customer to
get clarification.

COMPLETING
{
"status": "ok",
"id": "9c186090-4d39-11eb-83b9-096c29af",
"pictures": [
{
"name": "Shelves.jpg",
"result": [
{
"x": 779.7940717628705,
"y": 547.5179407176286,
"id": "IEEQ8SzBdspmbn37ursEh",
"type": "BoundingBoxes",
"label": "20.40",
"width": 83,
"height": 50,
"price_tag_id": 3,
"price_tag": 1,
"sku_id": 7
},
{
"x": 856.6479459178368,
"y": 88.98751950078005,
"id": "Kg-Zrgupgg19nkEoyk29e",
"type": "BoundingBoxes",
"label": "Mirinda",
"width": 34.66666666666663,
"height": 20,
"price_tag_id": 3,
"sku_id": 7,
"cap": 1
},
{
"x": 890.6479459178368,
"y": 97.65418616744671,
"id": "SKOO_fHYVqggjkOGCNhhl",
"type": "BoundingBoxes",
"label": "Mirinda",
"width": 42,
"height": 29.333333333333343,
"price_tag_id": 3,
"sku_id": 7,
"cap": 1
},
{
"x": 940.0440717628705,
"y": 332.4679407176284,
"id": "9VvX3_J58tW32pzMNmQL5",
"type": "BoundingBoxes",
"label": "Mirinda",
"width": 73.5,
"height": 62,
"price_tag_id": 3,
"sku_id": 7
},
{
"x": 878.4960998439942,
"y": 110.82776911076434,
"id": "FJz_P09njHWn8q1_URj8T",
"type": "BoundingBoxes",
"label": "Beverage",
"width": 158.4000000000001,
"height": 430.40000000000003,
"price_tag_id": 3,
"sku_id": 7
}
]
}
]
}
JSON response