umt won't work

[aweussom]

I followed the instructions at https://github.com/nathanrooy/rpi-urban-mobility-tracker

I'm getting this error:
(The first three lines as apparently just noise)

root@674b2ef1ec82:~# umt -video highway_01.mp4
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.summary API due to missing TensorBoard installation.
Traceback (most recent call last):
File "/usr/local/bin/umt", line 5, in
from umt.umt_main import main
File "/usr/local/lib/python3.7/dist-packages/umt/umt_main.py", line 15, in
from umt.umt_utils import parse_label_map
File "/usr/local/lib/python3.7/dist-packages/umt/umt_utils.py", line 26, in
encoder = gd.create_box_encoder(w_path, batch_size=1)
File "/usr/local/lib/python3.7/dist-packages/deep_sort_tools/generate_detections.py", line 123, in create_box_encoder
image_encoder = ImageEncoder(model_filename, input_name, output_name)
File "/usr/local/lib/python3.7/dist-packages/deep_sort_tools/generate_detections.py", line 97, in init
f"net/{input_name}:0")
File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/ops.py", line 3902, in get_tensor_by_name
return self.as_graph_element(name, allow_tensor=True, allow_operation=False)
File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/ops.py", line 3726, in as_graph_element
return self._as_graph_element_locked(obj, allow_tensor, allow_operation)
File "/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/ops.py", line 3768, in _as_graph_element_locked
"graph." % (repr(name), repr(op_name)))
KeyError: "The name 'net/images:0' refers to a Tensor which does not exist. The operation, 'net/images', does not exist in the graph."

Coral USB is working in the host os:

[40332.332886] usb 2-2: New USB device strings: Mfr=0, Product=0, SerialNumber=0

pi@pifem:~/coral/tflite/python/examples/classification $ python3 classify_image.py --model
models/mobilenet_v2_1.0_224_inat_bird_quant_edgetpu.tflite --labels models/inat_bird_labels.txt --input images/parrot.jpg
----INFERENCE TIME----
Note: The first inference on Edge TPU is slow because it includes loading the model into Edge TPU memory.
17.4ms
4.4ms
4.4ms
4.4ms
4.4ms
-------RESULTS--------
Ara macao (Scarlet Macaw): 0.77734

[aweussom]

Exactly the same problem in Ubuntu as in Raspberry PI OS/Docker.

[paulbozans]

I confirm, I have the same problem.

[Gazuru]

First you need to locate where your Python packages were installed, then in "deep_sort_tools/generate_detections.py" you need to change these two lines from :

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{input_name}:0")
self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{output_name}:0")

to this:

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"{input_name}:0")
self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"{output_name}:0")

[senker]

This is referred to the ubuntu installation I would assume, right? By making this change has everything worked correctly afterwards?

First you need to locate where your Python packages were installed, then in "deep_sort_tools/generate_detections.py" you need to change these two lines from :

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{input_name}:0")
self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{output_name}:0")

to this:

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"{input_name}:0")
self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"{output_name}:0")

[aweussom]

This is referred to the ubuntu installation I would assume, right? By making this change has everything worked correctly afterwards?

First you need to locate where your Python packages were installed, then in "deep_sort_tools/generate_detections.py" you need to change these two lines from :

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{input_name}:0")
self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{output_name}:0")

to this:

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"{input_name}:0")h

This is on Ubuntu also - however - the latest UMT code from GitHub have fixed this.

I have a NEW issue, bit I will create another thread for that :-)

[aweussom]

These two erranous lines are still there in the docker install.

[xuniluser]

This is referred to the ubuntu installation I would assume, right? By making this change has everything worked correctly afterwards?

First you need to locate where your Python packages were installed, then in "deep_sort_tools/generate_detections.py" you need to change these two lines from :

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{input_name}:0")
self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"net/{output_name}:0")

to this:

self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"{input_name}:0")
self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name(
     f"{output_name}:0")

This worked for me in RPi OS

[paulbozans]

In my case, yes. Get Outlook for Android<https://aka.ms/AAb9ysg>

…

________________________________ From: xuniluser ***@***.***> Sent: Monday, October 9, 2023 7:20:25 PM To: nathanrooy/rpi-urban-mobility-tracker ***@***.***> Cc: Paul Bozan ***@***.***>; Comment ***@***.***> Subject: Re: [nathanrooy/rpi-urban-mobility-tracker] umt won't work (#26) This is referred to the ubuntu installation I would assume, right? By making this change has everything worked correctly afterwards? First you need to locate where your Python packages were installed, then in "deep_sort_tools/generate_detections.py" you need to change these two lines from : self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name( f"net/{input_name}:0") self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name( f"net/{output_name}:0") to this: self.input_var = tf.compat.v1.get_default_graph().get_tensor_by_name( f"{input_name}:0") self.output_var = tf.compat.v1.get_default_graph().get_tensor_by_name( f"{output_name}:0") This worked for me in RPi OS — Reply to this email directly, view it on GitHub<#26 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ASFFA6JPMNGHTXCXD2OARKDX6QP4TAVCNFSM5EKWP6U2U5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCNZVGMZTANZXHA4A>. You are receiving this because you commented.Message ID: ***@***.***>

[jacksonalanhenry]

the deep_sort repo updated the version number from v1.2.0 to v1.3.0 and added the net for weighting in this commit
I'm going to submit a pull request on the Dockerfile so that this problem won't happen for other folks just looking to use the Dockerfile.

For tracking though, you can change this line:

RUN pip3 install git+https://github.com/mk-michal/deep_sort

to

RUN pip3 install git+https://github.com/mk-michal/deep_sort.git@v1.2.0