How can ı use tensorflow object detection api 2 saved_model.pb on jetson xavier NX?

Autonomous Machines Jetson & Embedded Systems Jetson Xavier NX
1

Hi, I trained ssd_mobilenet_v2_fpnlite_640x640_coco17_tpu-8 with tensorflow object detection api 2 and export saved_model.pb. Can ı use this model on jetson xavier or nano?

3

Hi,

Which inference frameworks do you prefer?

You can use TensorFlow directly on the XavierNX.
The package can be installed with the following instructions:

Or you can run it with TensorRT after you convert the model into ONNX format.
You can find an example for the TF object detection (TFOD) model below:

Thanks.

4

Thanks for reply. I tried to convert custom efficient-d0 model to tensorRT but I got this error.

Traceback (most recent call last):
File “create_onnx.py”, line 451, in
main(args)
File “create_onnx.py”, line 427, in main
effdet_gs.update_nms(args.nms_threshold, args.nms_detections)
File “create_onnx.py”, line 341, in update_nms
anchors_tensor = extract_anchors_tensor(box_net_split)
File “create_onnx.py”, line 284, in extract_anchors_tensor
anchors = np.concatenate([anchors_y, anchors_x, anchors_h, anchors_w], axis=2)
File “<array_function internals>”, line 6, in concatenate
ValueError: all the input array dimensions for the concatenation axis must match exactly, but along dimension 1, the array at index 0 has size 49104 and the array at index 2 has size 1

5

There is no update from you for a period, assuming this is not an issue any more.
Hence we are closing this topic. If need further support, please open a new one.
Thanks

Hi,

What kind of customization do you make?
Do you retrain the model with transfer learning? Or have modified the architecture?

Thanks.