Fever

Fever is a contactless fever thermometer with auto-aim. It combines a thermal camera with face detection from the Thermal Face project.

⚠️ This is an incomplete prototype. It is not a medical device. See issues for remaining work.

Parts

Part	Info	Price (2020)
FLIR Lepton 3.5	Thermal camera module (docs)	$199.00
PureThermal Mini	Thermal camera dev board (docs, library)	$99.00
BME680 Breakout Board	Ambient temperature, pressure, and humidity sensor (library)	$23.95
Raspberry Pi 4 Model B	Tiny computer (docs)	$35.00
Coral USB Accelerator	Edge TPU for faster face detection (docs, library)	$59.99

Setup

Image Raspbian and use sudo raspi-config:

Network Options > Wi-fi
Boot Options > Desktop / CLI > Console Autologin
Interfacing Options > SSH
Interfacing Options > SPI
Interfacing Options > I2C

Enable access to the thermal camera over UVC:

cd /tmp

git clone https://github.com/groupgets/libuvc
cd libuvc
mkdir build
cd build
cmake ..
make
sudo make install

sudo sh -c "echo 'SUBSYSTEMS==\"usb\", ATTRS{idVendor}==\"1e4e\", ATTRS{idProduct}==\"0100\", SYMLINK+=\"pt1\", GROUP=\"usb\", MODE=\"666\"' >> /etc/udev/rules.d/99-pt1.rules"

Install

cd

echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | sudo tee /etc/apt/sources.list.d/coral-edgetpu.list
curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -

sudo apt-get update
sudo apt-get install -y python3-venv python3-opencv
sudo apt-get install -y libedgetpu1-max python3-edgetpu
sudo apt-get install -y libatlas-base-dev libjasper-dev libhdf5-dev libqt4-dev
sudo apt-get install -y git

python3 -m venv venv
. venv/bin/activate
pip3 install --no-cache-dir tensorflow
pip3 install opencv-contrib-python
pip3 install numpy absl-py Pillow
pip3 install smbus2 bme680
pip3 install https://dl.google.com/coral/python/tflite_runtime-2.1.0.post1-cp37-cp37m-linux_armv7l.whl

git clone https://github.com/maxbbraun/fever.git
cd fever

curl -O https://github.com/maxbbraun/thermal-face/releases/latest/download/thermal_face_automl_edge_fast_edgetpu.tflite

Run

cd ~/fever

. venv/bin/activate
export LD_PRELOAD=/usr/lib/arm-linux-gnueabihf/libatomic.so.1
export PYTHONPATH=$PYTHONPATH:/usr/lib/python3/dist-packages

python fever.py --verbosity=1

Ambient temperature: 25 °C
Ambient pressure: 1010 hPa
Ambient humidity: 47 %
2 people
35 °C
34 °C
Frame took 83 ms (12.05 Hz)
Ambient temperature: 25 °C
Ambient pressure: 1010 hPa
Ambient humidity: 48 %
2 people
35 °C
34 °C
Frame took 87 ms (11.43 Hz)
...

Visualize

python fever.py --visualize

Flags

$ python fever.py --help

       USAGE: fever.py [flags]
flags:

fever.py:
  --[no]detect: Whether to run face detection.
    (default: 'true')
  --[no]display_metric: Whether to display metric units.
    (default: 'true')
  --face_confidence: The confidence threshold for face detection.
    (default: '0.5')
    (a number)
  --face_model: The TF Lite face detection model file compiled for Edge TPU.
    (default: 'thermal_face_automl_edge_fast_edgetpu.tflite')
  --max_num_faces: The maximum supported number of faces detected per frame.
    (default: '10')
    (an integer)
  --max_temperature: The maximum expected body temperature in centikelvin.
    (default: '31815')
    (an integer)
  --min_temperature: The minimum expected body temperature in centikelvin.
    (default: '29815')
    (an integer)
  --[no]visualize: Whether to visualize the thermal image.
    (default: 'false')

Try --helpfull to get a list of all flags.

maxbbraun / fever