Project Demo
Problem Statement
An empty IV bag can pose several issues, both medical and operational:- Extra pressure on the bag may exceed atmospheric within the bag, allowing air to continue to infuse into the patient once the bag is empty (if gravity fed). For the same reason, a bag that has been disconnected from the IV set should never be re-connected, as any extra air that may enter the flask could lead to embolism.
- Periodic monitoring by staff can be challenging depending upon staffing level and size of facility.
Solution
The prototype for this project will classify the IV fluid level into three categories:- Adequate
- Less than 50%
- Low level

edge-impulse-daemon
Now, the Portenta is connected to your Edge Impulse account.
For initial setup of the Portenta, follow the steps outlined here.
Dataset
I have collected 117 images of an IV fluid bag in various states, consisting of “adequate”, “less than 50%”, and “low level”.
Create Impulse:
In the “Create Impulse” section, configure the settings as shown here.
Neural Network
In “Neural Network”, I have used a single 2D convolution layer with flatten – DNN (10 neurons). The reason to go with one set of 2D Conv instead of more layers in 2D Conv, is memory consumption. Due to memory constraints on the device, I have one included one 2D Conv layer.

Deployment
In the “Deployment” section , select “Firmware – Portenta H7” and download it.
