Tensorflow Lite model loading error

developer0916 · December 29, 2022, 6:13am

I previously inquired about Tensorflow Lite model loading error.

I want to do object detection using the object detection api.
I was unable to convert and load a saved_model from sensorflow2 to Ttflite using BSP5.4 and TensorflowLite 2.3.1.
SSD MobileNet v2 320x320

github.com

tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md

# TensorFlow 2 Detection Model Zoo

[![TensorFlow 2.2](https://img.shields.io/badge/TensorFlow-2.2-FF6F00?logo=tensorflow)](https://github.com/tensorflow/tensorflow/releases/tag/v2.2.0)
[![Python 3.6](https://img.shields.io/badge/Python-3.6-3776AB)](https://www.python.org/downloads/release/python-360/)

<!-- mdlint off(URL_BAD_G3DOC_PATH) -->

We provide a collection of detection models pre-trained on the
[COCO 2017 dataset](http://cocodataset.org). These models can be useful for
out-of-the-box inference if you are interested in categories already in those
datasets. You can try it in our inference
[colab](../colab_tutorials/inference_tf2_colab.ipynb)

They are also useful for initializing your models when training on novel
datasets. You can try this out on our few-shot training
[colab](../colab_tutorials/eager_few_shot_od_training_tf2_colab.ipynb).

Please look at [this guide](running_on_mobile_tf2.md) for mobile inference.

<!-- mdlint on -->

This file has been truncated. show original

Tensorflow1 models could be converted to tflite and loaded. ssd_mobilenet_v1_quantized_coco

github.com

tensorflow/models/blob/master/research/object_detection/g3doc/tf1_detection_zoo.md

# TensorFlow 1 Detection Model Zoo

[![TensorFlow 1.15](https://img.shields.io/badge/TensorFlow-1.15-FF6F00?logo=tensorflow)](https://github.com/tensorflow/tensorflow/releases/tag/v1.15.0)
[![Python 3.6](https://img.shields.io/badge/Python-3.6-3776AB)](https://www.python.org/downloads/release/python-360/)

We provide a collection of detection models pre-trained on the
[COCO dataset](http://cocodataset.org), the
[Kitti dataset](http://www.cvlibs.net/datasets/kitti/), the
[Open Images dataset](https://storage.googleapis.com/openimages/web/index.html),
the [AVA v2.1 dataset](https://research.google.com/ava/) the
[iNaturalist Species Detection Dataset](https://github.com/visipedia/inat_comp/blob/master/2017/README.md#bounding-boxes)
and the
[Snapshot Serengeti Dataset](http://lila.science/datasets/snapshot-serengeti).
These models can be useful for out-of-the-box inference if you are interested in
categories already in those datasets. They are also useful for initializing your
models when training on novel datasets.

In the table below, we list each such pre-trained model including:

*   a model name that corresponds to a config file that was used to train this

This file has been truncated. show original

This query is below.
https://community.toradex.com/t/tensorflow-lite-model-loading-error/15295

This is the same content again.
Building Machine Learning Software with Reference Images for Yocto Project
I have now tried TensorflowLite2.8 with BSP5.7 based on the above. The conversion is also using Tensorflow2.8.
Tensorflow2 models can be loaded when running on CPU, but when using NPU, an error occurs. That is, whether or not /usr/lib/libvx_delegate.so is specified.

Why is this?
I am attaching the source code of the conversion, the source code of the execution, and log.
convert.py (1.9 KB)
log.txt (642 Bytes)
test.py (864 Bytes)
.

Best regards.

alvaro.tx · January 13, 2023, 4:02am

Dear @developer0916 , are you using Verdin iMX8M Plus, right? Because the initial post you mentioned used Apalis iMX8 that doesn’t have NPU.

EDIT: Nevermind, just read the log. We are checking but I’m reading here and there and I’m not 100% sure if model conversion from TensorFlow2 is supported as TensorFlow(1). Do you know?

Alvaro.

developer0916 · January 13, 2023, 5:21am

I am using Verdin-iMX8M.
Please close on this matter.
Resolved.
It appears that the model conversion was incorrect.
Tensorflow2 model also worked.

Best regards.

alvaro.tx · January 13, 2023, 5:48am

Thanks for the update. That is great to hear. Can you give a bit more information of what you changed to make it work for future reference?

Thanks,
Alvaro.

developer0916 · January 16, 2023, 1:25am

The following information was helpful

Armadillo page.

https://colab.research.google.com/github/atmark-techno/Armadillo-Base-OS-dev-guide-colab/blob/master/Armadillo_IoT_G4_create_detection_model.ipynb

Their manual is very well done.
I think we too need to create a manual for AI.
There is a demand for it.
I am in the process of doing some research for that.

Best regards.

alvaro.tx · January 16, 2023, 1:34am

Thanks for the info! Having a detailed manual would be great… If we have a Japanese manual, we will publish it ASAP.

ie. Linux OS カスタムマニュアル BSP5用

Alvaro.