Datasets:

Modalities:
Image
Languages:
Dutch
Libraries:
Datasets
License:
Search is not available for this dataset
image
imagewidth (px)
174
2k
label
class label
58 classes
00
00
00
00
00
00
11
11
11
11
11
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
11
11
11
11
122
122
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
11
11
11
11
00
00
00
00
00
00
00
00
11
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00
00

Vuurwerkverkenner

This dataset is used by the Vuurwerkverkenner, an application for linking snippets of exploded (heavy) fireworks to the type of firework that they originate from. You may find the application at www.vuurwerkverkenner.nl. The dataset consists of different types of fireworks that were investigated in casework by the Netherlands Forensic Institute.

Categories

If firework wrappers are very similar in their visual appearance, they may be grouped into categories. In most cases, a wrapper is not similar to any other wrapper, and a category simply consists of one unique wrapper.

Contents

The dataset consists of 185 categories, containing 332 unique wrappers in total. The dataset is organized as follows:

vuurwerkverkenner-data 
└───fireworks_0 
  └─── wrappers 
    └─── 0 
      └─── wrapper.jpg 
      └─── compleet exemplaar.jpg
      └─── gedemonteerd.jpg
    └─── 1 
      └─── wrapper.jpg 
      └─── compleet exemplaar.jpg 
      └─── gedemonteerd.jpg 
└───fireworks_1 
  └─── wrappers 
    └─── 0 
      └─── wrapper.jpg 
└───meta.json.gz 

The first level of folders (fireworks_0, fireworks_1, ...) corresponds to the category. The third level (0, 1, ...) corresponds to the wrappers within the category. For each wrapper, one or multiple images may be present. In each case, wrapper.jpg should be present, which is a scan of the full wrapper. Additionally, other images may be present as well, such as the entire fireworks article or a schematic drawing of its contents.

In meta.json.gz, we store both metadata and reference embeddings for the firework categories. It has the following structure:

{ 
  0: {
    "embeddings": [[...], [...], ...], 
    "wrappers": [ 
      0: {
        "wrapper_text": "abcdef", 
        "article_name": "abc",
        ... 
      }, 
      1: {...},
      2: {...} 
    ] 
  }, 
  1: {...} 
} 

The first level (meta[0], meta[1], meta[2] ...) indicates the fireworks category, which corresponds to the first level of the photo folders (i.e. meta[0] matches the folder fireworks_0).
Each category contains the reference embeddings and the metadata for each wrapper.
The third level, e.g. meta[0]['wrappers’][0], indicates the wrapper within the category and corresponds to the photo folder of that wrapper (e.g. fireworks_0/wrappers/0).

Reference embeddings

For each category, reference embeddings are provided. They are constructed by an AI model that is trained to create embeddings for (photos of) firework wrappers and snippets of exploded wrappers. When fed to this model, photos of snippets that belong to a given category should have similar embeddings to the ones for the wrappers. For more details about the model, see (here)[https://huggingface.co/NetherlandsForensicInstitute/vuurwerkverkenner].

Metadata

For each wrapper, metadata is provided. The fields are described below.

Field Description
text The full text that is on the wrapper (excluding non-Latin alphabets).
endangerment The level of danger or risk associated with the firework.
article_name The name of the item.
firework_type The type of firework.
firework_category The category or classification of the firework.
tube_length The length of the tube.
tube_diameter The diameter of the tube.
composition_burst_charge The composition of the main charge of the firework.
composition_other_mixtures The composition of other charges of the firework.
mass_burst_charge The mass of the main charge.
mass_other_mixtures The mass of other charges.
Downloads last month
39
Edit dataset card

Models trained or fine-tuned on NetherlandsForensicInstitute/vuurwerkverkenner-data