Configurable Watermark Rendering in DL Streamer - available soon!

[jangorecki1]

Dear All,
Configurable Watermark Rendering in DL Streamer will be available soon. This is long awaited feature.

More details here: HERE

The Challenge

• AI pipelines generate rich metadata (detections, labels, performance metrics), but visualizing them has been rigid and hard to control.
• Limited styling options made it difficult to adapt overlays for different use cases - debugging, demos, operations dashboards, or production deployments.

The Solution

• the enhanced gvawatermark introduces a configurable rendering layer via displ-cfg,
• This enables fine-grained control over how inference results are displayed - including labels, fonts, colors, thickness, background, and rendering device (CPU/GPU).

Benefits

• Flexible visualization - tailor overlays for different environments and audiences
• Configurability - adjust rendering without pipeline redesign
• Better operational clarity - cleaner, more readable AI insights
• Production-ready presentation

jan

[brmarkus]

Sounds great, looks great!

For various projects and demonstrators we also had used the "watermarks" plugins for e.g. blurring faces and anonymization of persons/bodies, using aaBB (at that time OBB were not (widely)used (at all)) and polygons (e.g. based on segmentation task) using different parameters (e.g. color, kernel, transparency, scaling-factor).
Would that be something for a future release?
What about allowing a pre- and/or post-processing "user defined function" (UDF) for custom operations, custom filters?

[brmarkus]

In the making: #618