Assertion Error when training BodyPoseNet with my custom data

• Network Type : BodyPoseNet

When I try to training BodyPoseNet with my custom data, the assertion error occurs. Before that, the first error is like this

tensorflow.python.framework.errors_impl.InvalidArgumentError: Dimensions must be equal, but are 180 and 64 for 'Loss/sub' (op: 'Sub') with input shapes: [8,180,102,19], [8,64,36,19].

So I think, I have to change target_shape in bpnet_train_m1_coco_custom.yaml which I wrote [64, 36].
But when I change it to [180,102], the assertion error occurs

INFO    2023-02-06 02:28:14,255| iva.common.logging.logging: Log file already exists at /workspace/tao-experiments/bpnet/models/custom_v1/status.json
Traceback (most recent call last):
  File "</usr/local/lib/python3.6/dist-packages/driveix/bpnet/scripts/train.py>", line 3, in <module>
  File "<frozen driveix.bpnet.scripts.train>", line 221, in <module>
  File "<frozen driveix.bpnet.scripts.train>", line 150, in main
  File "<frozen moduluspy.modulus.modulusobject.modulusobject>", line 158, in deserialize_maglev_object
  File "<frozen moduluspy.modulus.modulusobject.modulusobject>", line 145, in _deserialize_recursively
  File "<frozen moduluspy.modulus.modulusobject.modulusobject>", line 167, in deserialize_maglev_object
  File "<frozen moduluspy.modulus.modulusobject.modulusobject>", line 432, in wrapper
  File "<frozen driveix.bpnet.dataloaders.bpnet_dataloader>", line 150, in __init__
  File "<frozen driveix.bpnet.dataloaders.processors.label_processor>", line 57, in __init__
AssertionError
Telemetry data couldn't be sent, but the command ran successfully.
[WARNING]: <urlopen error [Errno -2] Name or service not known>
Execution status: FAIL

It only says “Assertion Error” so I don’t know what to do. Please help me to solve this problem. (Is there a possibility that the train-fold file was made incorrectly?)

From Body Pose Estimation — TAO Toolkit 4.0 documentation

  • Currently, BodyPoseNet only supports the given default skeleton configuration at pose_config_path. The inference pipelines do not support custom skeleton configuration at the moment.

Please check.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.