Documentation for Instance Segmentation Outputs

[pionex]

trafficstars

Is there any documentation for how to use the outputs of the instance seg model (exported to onnx) to get the typical bounding boxes, masks, etc?

I see 3 outputs:

name: dets tensor: float32[Concatdets_dim_0,Concatdets_dim_1,4]

name: labels tensor: float32[Addlabels_dim_0,Addlabels_dim_1,2]

name: 4245 tensor: float32[1,Addlabels_dim_1,108,108]

[isaacrob-roboflow]

Those are bounding boxes, label scores, and masks for the objects :) see our standardized ONNX / TensorRT benchmarking repo for an example pipeline https://github.com/roboflow/single_artifact_benchmarking/blob/main/sab/models/benchmark_rfdetr_seg.py