play-with-tflite

Repository for implementation Raspberry Pi + TensorFlow Lite to develop AI apps: Vehicle analytics.

Tools

Tested Hardware

• RasberryPi 4 Model B here, RAM: 4 GB and Processor 4-core @ 1.5 GHz
• microSD Card 64 GB
• 5M USB Retractable Clip 120 Degrees WebCam Web Wide-angle Camera Laptop U7 Mini or Raspi Camera

Tested Software

• OS Raspbian 10 (Buster) 32 bit armv7l, install on RasberriPi 4
• TensorFlow Lite library
• Python min. ver. 3.5 (3.7 recommended)

Getting Started

• Install TensorFlow Lite library (TensorFlow Lite APIs Python)

$ pip3 install https://github.com/google-coral/pycoral/releases/download/release-frogfish/tflite_runtime-2.5.0-cp37-cp37m-linux_armv7l.whl

Usage

Image Classification

$ git clone https://github.com/mheriyanto/play-with-tflite.git
$ cd play-with-tflite
$ cd examples
$ python3 classify.py --source /dev/video0 --model ../saved/models/mobilenet_v1_1.0_224_quant.tflite --labels ../saved/models/labels_mobilenet_quant_v1_224.txt

# Open on your browser and check http://0.0.0.0:5000/

Object Detection

$ python3 detection.py --source /dev/video0 --model ../saved/models/coco_ssd_mobilenet_v1_1.0_quant_2018_06_29.tflite --labels ../saved/models/coco_labels.txt

# Open on your browser and check http://0.0.0.0:5000/

Reference

• TensorFlow Lite Python classification example with Pi Camera: TensorFlow Lite example
• TensorFlow Lite Python object detection example with Pi Camera: TensorFlow Lite example
• Paper: MobileFaceNets: Efficient CNNs for Accurate Real-Time Face Verification on Mobile Devices
• Face Mask Detector (Tensorflow Lite): GitHub - tanhouren