Use Case and High-Level Description

This is a porosity weld recognition model. It runs on a video stream capturing welding tip and reports whether no welding is happening, produced weld is good or weld is porous.



Metric Value
Clip classification accuracy 97.14% (internal test set)
Temporal smoothing window size 16
GFlops 3.636
MParams 11.173
Source framework PyTorch*


Image, name: input, shape: 1, 3, 224, 224 in the format B, C, H, W, where:

  • B - batch size
  • C - number of channels
  • H - image height
  • W - image width

Expected color order is BGR.


The features is a blob with the shape 1, 3 containing logits for three output classes ("no weld", "normal weld" and "porosity").

Legal Information

[*] Other names and brands may be claimed as the property of others.