I use this Openpose’s Tensorflow version library https://github.com/ildoonet/tf-pose-estimation for Human pose detection. I like to run the model in Tensort for faster processing speed. CMU netowrk https://github.com/ildoonet/tf-pose-estimation/blob/master/tf_pose/network_cmu.py is quite straightforward and composed of conv, max_pool,and concat. All layers are supported in Tensorrt. The network has image is image input and output is concat_stage7.
When I run convert-to-uff, the output is
NOTE: UFF has been tested with TensorFlow 1.12.0. Other versions are not guaranteed to work
UFF Version 0.6.3
=== Automatically deduced input nodes ===
[name: "image"
op: "Placeholder"
attr {
key: "dtype"
value {
type: DT_FLOAT
}
}
attr {
key: "shape"
value {
shape {
dim {
size: -1
}
dim {
size: -1
}
dim {
size: -1
}
dim {
size: 3
}
}
}
}
]
=========================================
=== Automatically deduced output nodes ===
[name: "Openpose/concat_stage7"
op: "ConcatV2"
input: "Mconv7_stage6_L2/BiasAdd"
input: "Mconv7_stage6_L1/BiasAdd"
input: "Openpose/concat_stage7/axis"
attr {
key: "N"
value {
i: 2
}
}
attr {
key: "T"
value {
type: DT_FLOAT
}
}
attr {
key: "Tidx"
value {
type: DT_INT32
}
}
]
==========================================
Using output node Openpose/concat_stage7
Converting to UFF graph
No. nodes: 463
UFF Output written to cmu/cmu_openpose.uff
Is that ok in conversion?
But when I do inference the speed is same as running Tensorflow model.
Then Tensorrt output is (276000,) dimension.
But Tensorflow model has 4D tensor.
self.tensor_output = self.graph.get_tensor_by_name(‘TfPoseEstimator/Openpose/concat_stage7:0’)
(?, ?, ?, 57)
What could be wrong?