# Programming Q&A

**URL:** https://discuss.tinyml.seas.harvard.edu/c/programming-q-a/23.md

[Latest](https://discuss.tinyml.seas.harvard.edu/latest.md) · [Categories](https://discuss.tinyml.seas.harvard.edu/categories.md)

---

## [About the Programming Q&A category](https://discuss.tinyml.seas.harvard.edu/t/about-the-programming-q-a-category/377)

<div class="topic-metadata">

**Author:** [@vjreddi](https://discuss.tinyml.seas.harvard.edu/u/vjreddi)\
**Replies:** 0

</div>

Stuck on something? Ask for the community’s help. A place to ask your questions about programming tiny ML models and understanding how to deploy them on various devices.

---

## [VWW colab - looking to retrain the model](https://discuss.tinyml.seas.harvard.edu/t/vww-colab-looking-to-retrain-the-model/1617)

<div class="topic-metadata">

**Author:** [@desslok](https://discuss.tinyml.seas.harvard.edu/u/desslok)\
**Replies:** 0\
**Last updated:** [October 26, 2024, 6:29pm UTC](https://discuss.tinyml.seas.harvard.edu/t/vww-colab-looking-to-retrain-the-model/1617 "2024-10-26T18:29:18Z")

</div>

Is the colab for the pre-built VWW application available somewhere? I’d like to retrain the face detection for a different object. TIA

---

## [Cnn-trad-fpool3 model](https://discuss.tinyml.seas.harvard.edu/t/cnn-trad-fpool3-model/1583)

<div class="topic-metadata">

**Author:** [@moahmmedHamdy](https://discuss.tinyml.seas.harvard.edu/u/moahmmedHamdy)\
**Replies:** 0\
**Last updated:** [September 5, 2024, 3:59pm UTC](https://discuss.tinyml.seas.harvard.edu/t/cnn-trad-fpool3-model/1583 "2024-09-05T15:59:17Z")

</div>

Hi everyone, I’m looking to implement the CNN-trad-fpool3 model using TensorFlow 2.15, but I’m having difficulty finding any papers or resources that explain how to implement it in this version. Does anyone happen to ha…

---

## [Did something change about the TFlite converter?](https://discuss.tinyml.seas.harvard.edu/t/did-something-change-about-the-tflite-converter/1581)

<div class="topic-metadata">

**Author:** [@desslok](https://discuss.tinyml.seas.harvard.edu/u/desslok)\
**Replies:** 1\
**Last updated:** [September 4, 2024, 11:12pm UTC](https://discuss.tinyml.seas.harvard.edu/t/did-something-change-about-the-tflite-converter/1581 "2024-09-04T23:12:06Z")

</div>

From the course, there is the following colab: I caught the print(model.predict(np.array(\[10.0\]))) change needed, but now I’m wondering why the convert is breaking. tflite\_model = converter.convert()

---

## [How To Convert A Quantized Model to a C Header File](https://discuss.tinyml.seas.harvard.edu/t/how-to-convert-a-quantized-model-to-a-c-header-file/1500)

<div class="topic-metadata">

**Author:** [@Eric\_Van\_Jan](https://discuss.tinyml.seas.harvard.edu/u/Eric_Van_Jan)\
**Replies:** 0\
**Last updated:** [May 24, 2024, 3:20pm UTC](https://discuss.tinyml.seas.harvard.edu/t/how-to-convert-a-quantized-model-to-a-c-header-file/1500 "2024-05-24T15:20:46Z")

</div>

Hi all! I am a newbie to TinyML have built and quantized my model which classifies water as Safe or Unsafe for human consumption and would like to deploy it onto a Arduino Nano 33 BLE Sense board. I have searched a numbe…

---

## [AttributeError: module 'tensorflow.python.framework.graph\_util' during generating a model](https://discuss.tinyml.seas.harvard.edu/t/attributeerror-module-tensorflow-python-framework-graph-util-during-generating-a-model/1501)

<div class="topic-metadata">

**Author:** [@mw6059](https://discuss.tinyml.seas.harvard.edu/u/mw6059)\
**Replies:** 0\
**Last updated:** [May 24, 2024, 3:26pm UTC](https://discuss.tinyml.seas.harvard.edu/t/attributeerror-module-tensorflow-python-framework-graph-util-during-generating-a-model/1501 "2024-05-24T15:26:27Z")

</div>

I’m new at this. During Generate a TensorFlow Model for Inference I received the following error. AttributeError: module ‘tensorflow.python.framework.graph\_util’ has no attribute ‘convert\_variables\_to\_constants’ It lo…

---

## [Test microphone on Ardunino Nano 33 BLE Sense Lite](https://discuss.tinyml.seas.harvard.edu/t/test-microphone-on-ardunino-nano-33-ble-sense-lite/1496)

<div class="topic-metadata">

**Author:** [@mw6059](https://discuss.tinyml.seas.harvard.edu/u/mw6059)\
**Replies:** 0\
**Last updated:** [May 20, 2024, 10:50pm UTC](https://discuss.tinyml.seas.harvard.edu/t/test-microphone-on-ardunino-nano-33-ble-sense-lite/1496 "2024-05-20T22:50:52Z")

</div>

Reference: Course 3 (Deploying TinyML), Testing the Sensors. Hardware: TinyML Kit Software: tinyMLx - Built-in Microphone Test (from Examples → Harvard\_TinyMLx) Serial plotter seems to be hearing a lot of noise. Quite…

---

## [doubt on tinyML](https://discuss.tinyml.seas.harvard.edu/t/doubt-on-tinyml/1431)

<div class="topic-metadata">

**Author:** [@pree1998](https://discuss.tinyml.seas.harvard.edu/u/pree1998)\
**Replies:** 1\
**Last updated:** [December 31, 2023, 1:57am UTC](https://discuss.tinyml.seas.harvard.edu/t/doubt-on-tinyml/1431 "2023-12-31T01:57:36Z")

</div>

i have a machine learning model which recognise 4 different categories of gestures. and if any other gesture comes, how should it behave ?

---

## [In the visual wakewords example why is the number of samples 256 and not 512?](https://discuss.tinyml.seas.harvard.edu/t/in-the-visual-wakewords-example-why-is-the-number-of-samples-256-and-not-512/1293)

<div class="topic-metadata">

**Author:** [@timp](https://discuss.tinyml.seas.harvard.edu/u/timp)\
**Replies:** 0\
**Last updated:** [May 5, 2023, 2:53am UTC](https://discuss.tinyml.seas.harvard.edu/t/in-the-visual-wakewords-example-why-is-the-number-of-samples-256-and-not-512/1293 "2023-05-05T02:53:47Z")

</div>

In the visual wakewords, why was number of samples only 256 when we are reading 512 (bytes?) from the PDM? Is it because we need two bytes to make up one sample, but then i dont see where we put two bytes together to ma…

---

## [Problem Debugging : Keyword Detection Model](https://discuss.tinyml.seas.harvard.edu/t/problem-debugging-keyword-detection-model/1182)

<div class="topic-metadata">

**Author:** [@rutvijpathak](https://discuss.tinyml.seas.harvard.edu/u/rutvijpathak)\
**Replies:** 0\
**Last updated:** [October 31, 2022, 9:01am UTC](https://discuss.tinyml.seas.harvard.edu/t/problem-debugging-keyword-detection-model/1182 "2022-10-31T09:01:24Z")

</div>

Hi Everyone, I’m getting this error. Does anyone have tips for solving this issue? The results for recognition should contain 6 elements, but there are 4 in a 2-dimensional shape RecognizeCommands::ProcessLatestResul…

---

## [STN - Spatial Transformer Network](https://discuss.tinyml.seas.harvard.edu/t/stn-spatial-transformer-network/1089)

<div class="topic-metadata">

**Author:** [@vince](https://discuss.tinyml.seas.harvard.edu/u/vince)\
**Replies:** 0\
**Last updated:** [May 25, 2022, 10:22am UTC](https://discuss.tinyml.seas.harvard.edu/t/stn-spatial-transformer-network/1089 "2022-05-25T10:22:15Z")

</div>

Anyone got an STN working on tiny? They look like a CNN but a lot of float32 stuff going on, I was wondering if anyone had any success or problems

---

## [Using the tiny\_conv model for image classification tasks](https://discuss.tinyml.seas.harvard.edu/t/using-the-tiny-conv-model-for-image-classification-tasks/984)

<div class="topic-metadata">

**Author:** [@suburban-daredevil](https://discuss.tinyml.seas.harvard.edu/u/suburban-daredevil)\
**Replies:** 1\
**Last updated:** [April 16, 2022, 5:31pm UTC](https://discuss.tinyml.seas.harvard.edu/t/using-the-tiny-conv-model-for-image-classification-tasks/984 "2022-04-16T17:31:40Z")

</div>

I am trying to build a image classification model and deploy it onto the MCU. I have tried building my custom model and deploying. When the serial monitor is opened, it says that “Invoke() failed after Initialization fai…

---

## [Way to avoid low current shutoff](https://discuss.tinyml.seas.harvard.edu/t/way-to-avoid-low-current-shutoff/795)

<div class="topic-metadata">

**Author:** [@chanseok](https://discuss.tinyml.seas.harvard.edu/u/chanseok)\
**Replies:** 2\
**Last updated:** [July 12, 2021, 11:11am UTC](https://discuss.tinyml.seas.harvard.edu/t/way-to-avoid-low-current-shutoff/795 "2021-07-12T11:11:15Z")

</div>

Hi, Recently, I built a sort of framework to gather the data with BLE. While we remotely collect the data, the power bank is required. The problem is that the power bank is shutoff due to low current. After that, BLE i…

---

## [How to improve transfer learning/training?](https://discuss.tinyml.seas.harvard.edu/t/how-to-improve-transfer-learning-training/780)

<div class="topic-metadata">

**Author:** [@jhertz](https://discuss.tinyml.seas.harvard.edu/u/jhertz)\
**Replies:** 0\
**Last updated:** [July 6, 2021, 12:37am UTC](https://discuss.tinyml.seas.harvard.edu/t/how-to-improve-transfer-learning-training/780 "2021-07-06T00:37:39Z")

</div>

Hi all! Attached is a graph of my latest attempt at transfer learning from a mobile\_netV1 model already trained to 1 million steps, doing transfer learning with 50000 additional steps to my new dataset. The chart of trai…

---

## [Problem with quantization of GRU model](https://discuss.tinyml.seas.harvard.edu/t/problem-with-quantization-of-gru-model/759)

<div class="topic-metadata">

**Author:** [@yahia\_ma](https://discuss.tinyml.seas.harvard.edu/u/yahia_ma)\
**Replies:** 3\
**Last updated:** [June 29, 2021, 7:23am UTC](https://discuss.tinyml.seas.harvard.edu/t/problem-with-quantization-of-gru-model/759 "2021-06-29T07:23:14Z")

</div>

hii TinyML community, After a long search in other forums for my problem, any solution fund won’t work for me. So, I remember my membership in Tiny Machine Learning Community, I hope that you can help me to overcome thi…

---

## [Network architectures we could use with current version of TF Lite for Micro?](https://discuss.tinyml.seas.harvard.edu/t/network-architectures-we-could-use-with-current-version-of-tf-lite-for-micro/726)

<div class="topic-metadata">

**Author:** [@ramkoppu](https://discuss.tinyml.seas.harvard.edu/u/ramkoppu)\
**Replies:** 8\
**Last updated:** [June 27, 2021, 8:26am UTC](https://discuss.tinyml.seas.harvard.edu/t/network-architectures-we-could-use-with-current-version-of-tf-lite-for-micro/726 "2021-06-27T08:26:38Z")

</div>

In all the TinyML course examples, We were using either Fully connected layers or Tiny Convolution Network. What are the other network architectures we could use with current version of TF Lite for Micro ?

---

## [Need Help! Trying to combine Teachable Machine with Person Detection Script](https://discuss.tinyml.seas.harvard.edu/t/need-help-trying-to-combine-teachable-machine-with-person-detection-script/722)

<div class="topic-metadata">

**Author:** [@jhertz](https://discuss.tinyml.seas.harvard.edu/u/jhertz)\
**Replies:** 0\
**Last updated:** [May 28, 2021, 6:32pm UTC](https://discuss.tinyml.seas.harvard.edu/t/need-help-trying-to-combine-teachable-machine-with-person-detection-script/722 "2021-05-28T18:32:06Z")

</div>

I’m trying to use the TFLite Person Detection example project as a template and insert my own model that I’ve created using Teachable Machine. However, it’s not working as expected, and giving a “Invoke Failed()” error i…

---

## [Micro\_speech application "timing out"](https://discuss.tinyml.seas.harvard.edu/t/micro-speech-application-timing-out/552)

<div class="topic-metadata">

**Author:** [@stephenbaileymagic](https://discuss.tinyml.seas.harvard.edu/u/stephenbaileymagic)\
**Replies:** 3\
**Last updated:** [April 18, 2021, 6:26pm UTC](https://discuss.tinyml.seas.harvard.edu/t/micro-speech-application-timing-out/552 "2021-04-18T18:26:40Z")

</div>

Has anyone else see a timing out behavior in the micro\_speech example code? I have been leaving my model running and it has repeatedly stopped working after about 24 hours. Symptoms: -Device stops taking commands -Bui…

---

## [How do I capture images, so I can label them, and then train a ML model?](https://discuss.tinyml.seas.harvard.edu/t/how-do-i-capture-images-so-i-can-label-them-and-then-train-a-ml-model/618)

<div class="topic-metadata">

**Author:** [@halheinrich](https://discuss.tinyml.seas.harvard.edu/u/halheinrich)\
**Replies:** 4\
**Last updated:** [April 17, 2021, 1:48pm UTC](https://discuss.tinyml.seas.harvard.edu/t/how-do-i-capture-images-so-i-can-label-them-and-then-train-a-ml-model/618 "2021-04-17T13:48:20Z")

</div>

Ideally, I’d like to use the Arduino Nano 33 BLE Sense from the TinyML course to capture images. According to Pete Warden that’d be ‘difficult’. From the research I’ve done, it appears that using a webcam/dashcam a…

---

## [Tflite on MSP430](https://discuss.tinyml.seas.harvard.edu/t/tflite-on-msp430/586)

<div class="topic-metadata">

**Author:** [@premkumar1303](https://discuss.tinyml.seas.harvard.edu/u/premkumar1303)\
**Replies:** 4\
**Last updated:** [April 9, 2021, 2:36am UTC](https://discuss.tinyml.seas.harvard.edu/t/tflite-on-msp430/586 "2021-04-09T02:36:45Z")

</div>

I wanted to deploy a SimpleRNN model with one layer onto MSP430 microcontroller. Does MSP430 supports tflite? Is it possible to deploy neural networks onto MSP430?

---

## [PlatformIO and ESP32 development](https://discuss.tinyml.seas.harvard.edu/t/platformio-and-esp32-development/491)

<div class="topic-metadata">

**Author:** [@gameldar](https://discuss.tinyml.seas.harvard.edu/u/gameldar)\
**Replies:** 9\
**Last updated:** [March 25, 2021, 8:59pm UTC](https://discuss.tinyml.seas.harvard.edu/t/platformio-and-esp32-development/491 "2021-03-25T20:59:15Z")

</div>

Bringing this over from the projects topic Your reasons for switching to platformio were pretty similar to mine. I hit some serialization issues with the Arduino IDE but primarily I wanted a simple setup that I could u…

---

## [Dynamically updating a model on an embedded device](https://discuss.tinyml.seas.harvard.edu/t/dynamically-updating-a-model-on-an-embedded-device/395)

<div class="topic-metadata">

**Author:** [@Abu](https://discuss.tinyml.seas.harvard.edu/u/Abu)\
**Replies:** 9\
**Last updated:** [March 24, 2021, 12:56am UTC](https://discuss.tinyml.seas.harvard.edu/t/dynamically-updating-a-model-on-an-embedded-device/395 "2021-03-24T00:56:10Z")

</div>

I was wondering about updating the neural network model already deployed on device without compiling and uploading the whole application. Since we will have better performing models we would want to replace the older mo…

---

## [Arduino Nano BLE write value to a peripheral](https://discuss.tinyml.seas.harvard.edu/t/arduino-nano-ble-write-value-to-a-peripheral/496)

<div class="topic-metadata">

**Author:** [@little\_wing](https://discuss.tinyml.seas.harvard.edu/u/little_wing)\
**Replies:** 4\
**Last updated:** [March 17, 2021, 7:52pm UTC](https://discuss.tinyml.seas.harvard.edu/t/arduino-nano-ble-write-value-to-a-peripheral/496 "2021-03-17T19:52:46Z")

</div>

Not currently ML (although it will be) I’m trying to write a value to a peripheral (Hue bulb) using BLE on the Arduino Nano. So far I can connect to the peripheral using the nano but when I try to write (or read) a val…

---

## [Micro speech example from TensorFlow on an ESP32](https://discuss.tinyml.seas.harvard.edu/t/micro-speech-example-from-tensorflow-on-an-esp32/376)

<div class="topic-metadata">

**Author:** [@gameldar](https://discuss.tinyml.seas.harvard.edu/u/gameldar)\
**Replies:** 6\
**Last updated:** [February 22, 2021, 4:09pm UTC](https://discuss.tinyml.seas.harvard.edu/t/micro-speech-example-from-tensorflow-on-an-esp32/376 "2021-02-22T16:09:40Z")

</div>

Is there a good place to ask technical/implementation questions related to tiny ml - e.g. something like the programming questions forum of Arduino (Programming Questions) to get assistance with troubles implementing spe…
