Owlish

Demo

• View the demo here

Quick Start without BLE from RSL-10 Sense Kit

• Make sure to install all dependencies with pip3 as the tensorflow lite can only run on Python Version 3.5, 3.6 & 3.7 only.
• Replace credentials for Cloudinary and Twilio Whatsapp API (in upload_cloudinary.py & send_twilio.py) with your own.
• Clone and install the Tensorflow lite object detection model example by following the README.md in their official repository
• Once cloned and installed,cd to examples/lite/examples/object_detection/raspberry_pi/ and insert the following files from this repository (owlish_run.py, send_twilio.py, upload_cloudinary.py & detect_picamera.py).
• In owlish_run.py replace activate_status variable to True to run without activation from the BLE device and comment out scan_bluetooth.start() in the main.
• Run the application with python3 owlish_run.py

Quick Start using BLE from RSL-10 Sense Kit

• The only difference is to use sudo pip3 for installing every dependency (including the tensorflow lite) as we'll need to expose modules to the root user. This is because running the owlish_run.py requires root permissions for using the Bluetooth Peripheral on the PI.
• Follow the rest of the steps in Quick Start without BLE from RSL-10 Sense Kit without modifying owlish_run.py.
• Run the application with sudo python3 owlish_run.py

Screenshot

• Once you have the application up and running, you should receive something similar!

Built using

Hardware components

• Raspberry Pi 3B+
• A Pi Camera Module V2
• RSL10-SENSE-GEVK dev kit

Software components

• Twilio WhatsApp
• TensorFlow lite
• Bluepy
• Cloudinary

Misc

If you'd like to learn more, I did a technical write up over at dev.to

Future Improvements & Todo's

• Using a personal smartphone to activate the system when you're away instead of using the RSL10-SENSE-GEVK dev. This would be a better option but requires an additional WiFi adapter dongle.
• Using Twilio Autopilot for handoff to relevant authorities.
• Boost inference performance with Coral USB accelerator
• Dockerize the application

License

This project is licensed under MIT.