No active modelSign in
About the project

A closer look at structural surfaces.

Deep-learning segmentation meets practical inspection tools.

CrackVision AI

Built for the detail between pixels.

CrackVision AI is a workspace for applying deep-learning semantic segmentation to structural surface crack inspection. It brings trained-model management, image and video processing, geometric analysis, and downloadable reports into one inspection workflow.

The intended CrackSegNet architecture combines a lightweight encoder, residual ASPP for multi-scale context, attention mechanisms including CBAM, and a decoder producing a binary segmentation output. The exact trained implementation must be supplied to ensure checkpoint compatibility.

Research tools, transparent results.

Segmentation quality depends on the trained model and input conditions. This application makes no unsupported accuracy claims. Geometry is reported in original-image pixels, and the results should be interpreted within the context of the inspection.

Semantic segmentationResidual ASPPCBAM attentionConnected componentsPyTorchKeras

From your model to your report.

Bring the checkpoint you trained, inspect the original and predicted masks side by side, and keep the complete analysis in your own account.

Open workstation
© 2026 CrackVision AI / Surface intelligenceImage-space measurements. Engineering context matters.