search

bitsy-ai / rpi-object-tracking

Object tracking tutorial using TensorFlow / TensorFlow Lite, Raspberry Pi, Pi Camera, and a Pimoroni Pan-Tilt Hat.
https://medium.com/@grepLeigh/real-time-object-tracking-with-tensorflow-raspberry-pi-and-pan-tilt-hat-2aeaef47e134
MIT License
184 stars 68 forks source link

Multiple Issues ... #52

Open iCUE-Solutions opened 3 years ago

iCUE-Solutions commented 3 years ago

EDIT: 12/30/2020 Created the entire SD from scratch.

Turns out: Allocation error is only a warning. So, ignore it. Biggest issue is the missing TensorFlow wheel. I used this description to address the issue.

https://qengineering.eu/install-tensorflow-2.2.0-on-raspberry-pi-4.html

The following line only works without the <sudo -H> sudo -H pip3 install tensorflow-2.2.0-cp37-cp37m-linux_armv7l.whl

Now most of the cases work, except the tracker always goes off into never never land. Will keep you posted.

I was so excited when I found this demo and still you did a wonderful job! It's just that I struggle to get the parts to work.

It seems since you started this project there has been many updates in many areas and nothing really matches anymore e.g. description in the blog and various versions of code and frameworks. Sorry, just my best guess. I'm new to raspi and ML on raspi with CORAL -- latter tech is cool but it appears more demos don't work for a newbie.

Hours in the only thing that works according to script is : rpi-deep-pantilt test pantilt. The thing moves! Yes.

As for the next one : rpi-deep-pantilt test camera it's not crashing but I also can't see anyting. Despite the fact that I have turned on raspi cam and rebooted .... I see the desktop on my TV.

rpi-deep-pantilt detect --edge-tpu seems to load the labels (shows is it but ...) but presents this afterwards. I guess some error

EDIT: investigated more and the error below only happens after CPU mem is at least set to 64MB. Prior to this things crash with different issues.

2020-11-29 16:36:18.764478: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 307200 exceeds 10% of free system memory. 2020-11-29 16:36:22.000636: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 307200 exceeds 10% of free system memory. 2020-11-29 16:36:22.023519: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 307200 exceeds 10% of free system memory. 2020-11-29 16:36:22.072725: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 307200 exceeds 10% of free system memory. 2020-11-29 16:36:22.093627: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 307200 exceeds 10% of free system memory.

Now, this is only after I installed the latest ftlite framework from the Coral website. Prior to this there were other issues.

I understand you are doing this in your spare time and I truly appreciate you sharing this. Please don't take the text above as a rant. By no means is it meant that way.

I'm just wondering if there is some help you can offer on how to check if the correct libs are in place .... and how to make this work. I'm almost 3 days in with fiddling and now want to see it work.

Thanks in advance, Dirk

p.s. your blog still says: pip install https://github.com/leigh-johnson/Tensorflow-bin/blob/master/tensorflow-2.0.0-cp37-cp37m-linux_armv7l.whl\?raw\=true

which leads into this error:

Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple Collecting tensorflow==2.0.0 ERROR: HTTP error 404 while getting https://github.com/leigh-johnson/Tensorflow-bin/blob/master/tensorflow-2.0.0-cp37-cp37m-linux_armv7l.whl?raw=true ERROR: Could not install requirement tensorflow==2.0.0 from https://github.com/leigh-johnson/Tensorflow-bin/blob/master/tensorflow-2.0.0-cp37-cp37m-linux_armv7l.whl?raw=true because of HTTP error 404 Client Error: Not Found for url: https://github.com/bitsy-ai/tensorflow-arm-bin/blob/main/tensorflow-2.0.0-cp37-cp37m-linux_armv7l.whl for URL https://github.com/leigh-johnson/Tensorflow-bin/blob/master/tensorflow-2.0.0-cp37-cp37m-linux_armv7l.whl?raw=true

Martin2kid commented 3 years ago

Moehrenbaum,

You are correct that 10% CPU memory is just warning & it has nothing to do with rpi-deep-pantilt and you can ignore that.

I'm newbie too, obviously you saw Legh's post at Toward Data Science as you stated "pip install https://github.com/leigh-johnson/Tensorflow-bin/blob/master/tensorflow-2.0.0-cp37-cp37m-linux_armv7l.whl\?raw\=true".

You are leaving this message at Github that actually specify "$ pip install https://github.com/leigh-johnson/Tensorflow-bin/releases/download/v2.2.0/tensorflow-2.2.0-cp37-cp37m-linux_armv7l.whl" which I don't think it is same version as you installed(Leigh stated it is "community-built wheel" which imply it is special version).

I also noticed it takes multiple attempts to Pip install that file but I was able to install in 4th try about few weeks ago.

If you are using Coral USB, please follow in step 12 in Toward Data Science blog.

I've also noticed that discrepancy between Toward & Github post but I must still give her all the credit for sharing this code with us and it is up to us carefully try one post after another or combination and further reading through all available issue posted including closed post.

Cheers!

iCUE-Solutions commented 3 years ago

Moehrenbaum,

You are correct that 10% CPU memory is just warning & it has nothing to do with rpi-deep-pantilt and you can ignore that.

I'm newbie too, obviously you saw Legh's post at Toward Data Science as you stated "pip install https://github.com/leigh-johnson/Tensorflow-bin/blob/master/tensorflow-2.0.0-cp37-cp37m-linux_armv7l.whl\?raw\=true".

You are leaving this message at Github that actually specify "$ pip install https://github.com/leigh-johnson/Tensorflow-bin/releases/download/v2.2.0/tensorflow-2.2.0-cp37-cp37m-linux_armv7l.whl" which I don't think it is same version as you installed(Leigh stated it is "community-built wheel" which imply it is special version).

I also noticed it takes multiple attempts to Pip install that file but I was able to install in 4th try about few weeks ago.

If you are using Coral USB, please follow in step 12 in Toward Data Science blog.

I've also noticed that discrepancy between Toward & Github post but I must still give her all the credit for sharing this code with us and it is up to us carefully try one post after another or combination and further reading through all available issue posted including closed post.

Cheers!

Martin2kid, thanks again for helping out. I got it all working. I guess I simplified to my skill level :) AND a big thanks to Leigh for sharing her awesome work.