עברית

Train mask image classifier model in Google Colab (with GPU)

• Based on MobileNet V2 (SSD) pre-trained model:

  (Download TFLite model after training)

Install Raspberry Pi OS 64-bit on RPi 4B

Download from here

Optional - before installing, add an empty file called ssh (without extension) to the root of the drive, to enable SSH.

Overclock Raspberry Pi to 1750MHz, overvolt to 3V

sudo nano /boot/config.txt

arm_freq=1750
over_voltage=3

Optional - enable these lines on /boot/config.txt for VNC:

hdmi_force_hotplug=1
disable_overscan=1
gpu_mem=128

Post install

After install, in terminal:

sudo apt update
sudo apt full-upgrade

Istall OpenCV v4.5.0 or higher (for arch64)

Q-engineering tutorial

Install TensorFlow 2 v2.4.0 or higher (for arch64)

Q-engineering tutorial

Ater install, in terminal:

sudo apt install protobuf
pip install pycocotools
pip install tf_slim
pip install tensorflow_hub

install protos for TensorFlow 2 Object Detection API

In terminal:

chmod +x install-requirements.sh
./install-requirements.sh

Clone repo

In terminal:

git clone --recurse-submodules https://github.com/lgariv/CollegeProject.git

Download TFLite model from Google Colab and place in the same folder

Run

In terminal:

python Object_detection_webcam_tflite.py