Skip to content

Latest commit

 

History

History
93 lines (61 loc) · 3.23 KB

README.md

File metadata and controls

93 lines (61 loc) · 3.23 KB
Raw

Micropython implementation of the Person Detection Example

Model Architecture

Model Properties

Upstream example

https://github.com/tensorflow/tflite-micro/blob/main/tensorflow/lite/micro/examples/person_detection/README.md

Training the model

https://github.com/tensorflow/tflite-micro/blob/main/tensorflow/lite/micro/examples/person_detection/training_a_model.md

Files

person_detect_model.tflite

I was able to restore the tflite model from the person_detection_int8 archive. It was a .cc file but I converted it back into a binary file and verified the architecture in netron.

person_image_data.dat

I was able to create this (96,96) binary file from the corresponding .cc file from the person_detection_int8 archive.

person image

no_person_image_data.dat

I was able to create this (96,96) binary file from the corresponding .cc file from the person_detection_int8 archive.

not a person

show-test-images.py

A script that runs in python3 to load and display the two test images.

Running in unix port

Download the latest build of the micropython unix port:

  1. Click on "Actions" tab.
  2. Click on "Select Workflow" button.
  3. Choose "UNIX"
  4. Click on the commit message name for the most recent successful build
  5. Click on microlite-unix-micropython in the "Artifacts" area to download a zip file containing the micropython command for unix. It can also run within the windows subsystem for linux.

Run micropython /examples/person_detection/person_detection.py

  ~/git/tensorflow-micropython-examples/examples/person_detection$ ./micropython ./person_detection.py
interpreter_make_new: model size = 300568, tensor area = 139264
Classify No Person Image
setup 9612 bytes on the inputTensor.
'not a person' = -25, 'person' = 25
Classify Person Image
setup 9612 bytes on the inputTensor.
'not a person' = -113, 'person' = 113

Running on ESP32 port without a camera

NOTE: The person detection model is 300 kb so you need to use a board with SPI RAM

Download the latest build of the micropython microlite spiram board for the esp32 port:

  1. Click on "Actions" tab.
  2. Click on "Select Workflow" button.
  3. Choose "ESP32"
  4. Click on the commit message name for the most recent successful build
  5. Click on microlite-spiram-16m-esp32-firmware in the "Artifacts" area to download a zip file containing the esp32 firmware you can now flash.

After flashing upload these files from here onto the board:

  1. no_person_image_data.dat
  2. person_image_data.dat
  3. person_detect_model.tflite
  4. person_detection.py

Then import person_detection to run the example: ESP32 Running Person Detection

Running on an ESP32 With a Camera

This has been tested using an ESP32-CAM-MB and an M5 Timer Camera using the MICROLITE_SPIRAM_CAM Firmware.

Flash that firmware and then copy the esp32-cam/person_detection_cam.py and person_detect_model.tflite model to the camera.

Run import person_detection_cam to activate the demo program. The led will illuminate when it thinks a person is in frame.