Converting TensorFlow 2 Models to TensorFlow Lite and TFLite Object Detection on the Raspberry Pi

Timestamps
0:00 Intro
1:14 TensorFlow Lite Conversion
10:46 Object Detection on the Raspberry Pi
___________________________________________________________________________
Happy Thanksgiving! It's been a while since I last posted, but by popular demand, here's a guide on TensorFlow Lite Conversion! If you have any issues, feel free to let me know in the comments or raise an issue in GitHub. You can contact me directly through Discord. My handle is Troy and Abed#6396. To not miss out on future content coming soon, make sure to subscribe and turn on notifications. If you enjoyed the video, make sure to like. If you didn't, feedback would also be appreciated so I can improve the quality of my videos! :)

___________________________________________________________________________
I DO NOT OWN OR TAKE CREDIT FOR THE OUTRO MUSIC IN THE VIDEO
Original Soundtrack (OST) from the anime Made In Abyss (2017)
Music composed by Kevin Penkin
Style: Original Soundtrack
___________________________________________________________________________

[gihanna]

Thank you! You provide the most detailed and helpful videos!

[necatikirilmis]

Thank you so much bro! this exam helped a lot :)

[geraldmusandirire]

Do you know how I can get the coordinates of the bounding boxes in real time from your code? I need to use them for something else. Thank you

[renovationfox]

Hi, thanks for the tutorial. When running the custom model I get following TypeError--> only size-1 arrays can be converted to python scalars. Do you know how to fix this? Thanks

[tomrabhishek]

Hey, i have .ipynb file, i wanna use it in flutter app, can you please tell me what should i do

[anthonyoluyele]

Thank you Armaan, while trying to use my custom model. I am having this error:

ValueError: Cannot set tensor: Dimension mismatch. Got 3 but expected 1 for dimension 3 of input 0.

[fadilahduljeet]

Thank you, No metadata is displayed when i open it on android studio ? Why ?

[anirudhatadapatri]

Armaan, while I am running any of the python scripts written for testing out the model on Pi, I'm getting "TypeError: numpy.float32 has no len()". This is in the line 142 of pi camera object detection script. What can be the reason and I'd be very grateful if you could help me out with this. Thanks in advance

[karimbensaid]

i have this error :buffer deduplication procedure will be skipped when flatbuffer IS not properly loaded but the file model.tflite created

[gwizikd]

Thank you for the good lecture video.
While I was learning one by one while watching the existing video, I had a question, so I left a comment like this.
I practiced doing custom models on YouTube the other day and checked it with a computer. I did convert this to a tflite file in YouTube this time, but can't I do object-detection with this?

[karimbensaid]

hey model.tflite size is bigger about 11293 ko Can you help me?

[juniorhernandez]

Thank you so much for the tutorial. I got an error when installing prerequisites, exactly when installing tflite_runtime. The error: is not a supported wheel on this platform." I have a raspberry pi 3b+. Which python version are you using? Thanks for your help!

[muhammadyoga]

hello can you help me ? convert my script tensorflow to tensorflow lite?

[seifajimi]

can you please help me, i got segmentation fault at the end.

[erfansn]

Hello, thank you for the tutorial, but I have trouble converting to tflite and it does not complete, and I think because of Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found can you help me convert my model thank you

[lionking]

Hello, Thanks for this tutorial, but when I ran the python script to convert to tflite the model.tflite is 1kb. what should I do to fix this? my tf version ended up being 2.5.0-dev20201211 so slightly different from yours.

[kartikamerdekawatim]

what's type a model object detection that u use? yolo/ssd/f-rcnn? thankyouu

[AbdulHaseeb-ejgk]

Hey Armaan, when I'm running TFLite-PiCamera-OD.py
I'm getting the following error
for i in range (len(scores)):
TypeError: object of type 'numpy.float32' has no len()

[juanchin]

any advices on getting more frames per second?

[dropd]

Hi, thanks for the tutorial.Can you make real time license plate recognition system with TensorFlow and Raspberry Pi?