Deepness QGIS plugin demonstration

Deepness for QGIS – deep learning, no GPU needed This video shows Marek Kraft putting his Deepness plugin through its paces: load an orthophoto, pick a model from the built-in zoo, run inference on the CPU, and view the results as standard QGIS layers. After a car-detection pass he switches to building-footprint segmentation, tweaks styling, and exports the masks – all inside the QGIS interface. The demo pairs with a conversation where Marek dives into development details, real-world use cases and the growing user community. PROJECT LINK • Docs & download – https://qgis-plugin-deepness.readthedocs.io/en/latest/ 🚀 TIMELINE 00:00 Intro and plugin overview 00:10 Selecting the input imagery layer 00:40 Choosing a pretrained car-detection model from the model zoo 01:25 Running inference on the sample scene 01:40 Results: 92 cars detected with bounding boxes 02:15 Scaling up – city-wide car counting on CPU 03:50 Switching to a UNet-based building-footprint segmentation model 04:20 Generating segmentation masks and polygons 05:10 Styling layers and colour tweaks in QGIS 05:45 Exporting results as GeoTIFF and other formats 06:20 Wrap-up and future roadmap KEY TAKE-AWAYS • CPU-friendly: runs on any machine; a GPU only speeds things up. • Model zoo: download ready-to-use detection and segmentation models or load your own ONNX files. • Seamless integration: inference outputs appear as normal QGIS layers you can style, query and export. • Scales up: tile-overlap and resolution controls support neighbourhood- to city-scale analysis. • Active community: new models and workflows are being added all the time. Enjoy the demo? Try the plugin, star the repo, and tell us what you build!