Jetson Nano Detection and Tracking

Installing and setting up the new Nvidia Jetson Nano was surprisingly time consuming and unintuitive. From protobuf version conflicts, to Tensorflow versions, OpenCV recompiling with GPU, models running, models optimized, and general chaos in the ranks.

This repository is my set of install tools to get the Nano up and running with a convincing and scalable demo for robot-centric uses.

2 Likes

Nice, will keep an eye on it. Having some trouble myself.

I was able to modify this to get an rtsp feed working from a wyze cam. Thanks again for this, save me countless hours.

Awesome! I’m glad you got some use out of it. I’m curious, what’s your application?

Best!
Steve

Detect presence of community cat to control small outdoor fan during the summer.

That’s super awesome! Glad that I could help.

Gonna give this a try

The algorithm works perfectly.
How can I make it work with an IP camera?
I view my IP camera with:
cap = cv2.VideoCapture (“rtsp: // admin: admin@192.168.1.130: 554 / cam / realmonitor? channel = 1 & subtype = 1”).

You can change this line here with your new capture mode:

Thank you very much for your previous response.
Now I work with the IP camera at the beginning, and then the following came out:
Debug: Running at: 8.329104874774611 Hz.
Debug: Found objects: person.
Debug: Running at: 9.681494082806092 Hz.
[NULL @ 0x2bc0c6f0] missing picture in access unit
Debug: Found objects: person.
Debug: Running at: 8.944051485342756 Hz.
Debug: Found objects: person.
Debug: Running at: 9.135469144434062 Hz.
Debug: Found objects: person.
Debug: Running at: 8.874504891837908 Hz.
Debug: Found objects: person.
Debug: Running at: 9.270204442479832 Hz.
Failed to capture frame!
Traceback (most recent call last):
File “jetson_live_object_detection.py”, line 89, in
live_detection.start()
File “jetson_live_object_detection.py”, line 59, in start
scores, boxes, classes, num_detections = self.detector.detect(img)
File “/home/dlinano/jetson_nano_detection_and_tracking/src/object_detector.py”, line 24, in detect
scores, boxes, classes, num_detections = self.tf_sess.run(self.tf_tensors, feed_dict={self.tf_input: img_expanded})
File “/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py”, line 929, in run
run_metadata_ptr)
File “/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py”, line 1121, in _run
np_val = np.asarray(subfeed_val, dtype=subfeed_dtype)
File “/usr/local/lib/python3.6/dist-packages/numpy/core/_asarray.py”, line 85, in asarray
return array(a, dtype, copy=False, order=order)
TypeError: int() argument must be a string, a bytes-like object or a number, not ‘NoneType’

Thanks for your guide.
This worked for me, as I underline here: https://github.com/JordanMicahBennett/live_ai_object-detection-on-tiny-jetson-neural-nano-computer

Hi there,

I am trying to run this on Ubuntu 18.04, with the Jetpack 4.4 & Deepstream SDK 5.0. Everything installed correctly up until I try run the models. Once I do I get this error:

"ImportError: libcublas.so.10.0: cannot open shared object file: No such file or directory

Failed to load the native TensorFlow runtime.

See https://www.tensorflow.org/install/errors"

Anyone have a clue why? Thanks!

1 Like