ip-camera-image-inference | Python script using machine learning to classify images

ip-camera-image-inference is a Python library typically used in Internet of Things (IoT) applications. ip-camera-image-inference has no bugs, it has no vulnerabilities, it has a Permissive License and it has low support. However ip-camera-image-inference build file is not available. You can download it from GitHub.

This Python script is meant to be executed as a systemd service. It will subscribe an MQTT broker to receive JPEG images from an IP camera. The machine learning framework TensorFlow is used to classify images. A deep convolutional neural network (CNN) based on the Inception v3 model needs to be retrained with a set of images from your IP camera to classify the image either as person or as nothing. If a person is detected in the image, the Python script will publish the image to a specified MQTT topic. A threshold can be configured for this to publish even images with a lower probability score. This script was designed to work on a Raspberry Pi 3 (ARM architecture), but it also runs on x86_64 and should actually run on all platforms where TensorFlow and Python are available. On a Raspberry Pi 3 this is pretty slow (roughly three seconds per image). So make sure you have a good pre-processing of your IP camera feed and deliver only images where potentially movement was detected. You could either use the built-in motion detection of your IP camera or preferable a third party software detection like Motion for this.