July 13, 2026 Dataset Open Access

Radoynova, Martina; Pantze, Samuel; De, Trina; Günther, Ulrik; Yakimovich, Artur

Citation Style Language JSON Export

{
  "title": "Metric-Guided Synthetic Image Data Rendering for Deep Learning compatible with Agentic AI", 
  "author": [
    {
      "family": "Radoynova, Martina"
    }, 
    {
      "family": "Pantze, Samuel"
    }, 
    {
      "family": "De, Trina"
    }, 
    {
      "family": "G\u00fcnther, Ulrik"
    }, 
    {
      "family": "Yakimovich, Artur"
    }
  ], 
  "publisher": "Rodare", 
  "issued": {
    "date-parts": [
      [
        2026, 
        7, 
        13
      ]
    ]
  }, 
  "id": "4780", 
  "version": "Version 1", 
  "DOI": "10.14278/rodare.4780", 
  "type": "dataset", 
  "language": "eng", 
  "abstract": "<p>This repository contains the datasets, 3D rendering scene files, and trained model weights accompanying the manuscript <strong>&quot;Metric-Guided Synthetic Image Data Rendering for Deep Learning compatible with Agentic AI.&quot;</strong> The deposited files provide the complete procedural pipeline and resulting artefacts used to evaluate zero-shot instance segmentation of virological plaque assays (based on the <a href=\"https://rodare.hzdr.de/record/3003\">VACVPlaque dataset</a>).</p>\n\n<p>Repository Contents</p>\n\n<ul>\n\t<li>\n\t<p><strong><code>datasets/</code></strong> Contains the synthetic image datasets generated at three varying levels of perceived realism (High, Medium, Low) and in two sizes (Small [S]: 100 images; Large [L]: 1,000 images). It also includes the <code>Mix</code> dataset, which consists of 90 high-realism synthetic images and 10 real images.</p>\n\t</li>\n\t<li>\n\t<p><strong><code>blender/</code></strong> Includes the procedural 3D scene files used to generate the synthetic datasets via Blender. These files can be used manually or coupled with the SynthClaw agentic skill for automated rendering.</p>\n\n\t<ul>\n\t\t<li>\n\t\t<p><code>Scene_High.blend</code></p>\n\t\t</li>\n\t\t<li>\n\t\t<p><code>Scene_Medium.blend</code></p>\n\t\t</li>\n\t\t<li>\n\t\t<p><code>Scene_Low.blend</code></p>\n\t\t</li>\n\t</ul>\n\t</li>\n\t<li>\n\t<p><strong><code>model_weights/</code></strong> Contains the trained single-shot architecture model weights and threshold configurations used to evaluate the zero-shot performance of each generated dataset.</p>\n\n\t<ul>\n\t\t<li>\n\t\t<p><strong>Evaluated Conditions:</strong> <code>High_L</code>, <code>High_S</code>, <code>Medium_L</code>, <code>Medium_S</code>, <code>Low_L</code>, <code>Low_S</code>, and <code>Mix</code>.</p>\n\t\t</li>\n\t\t<li>\n\t\t<p><strong>Files per Condition:</strong> Each subdirectory includes the best and last model weights (<code>weights_best.h5</code>, <code>weights_last.h5</code>) alongside the corresponding threshold optimization parameters (<code>thresholds1.json</code>, <code>thresholds2.json</code>) for large and small object segmentation.</p>\n\t\t</li>\n\t</ul>\n\t</li>\n</ul>"
}