We are devoting €10.000 euros for open source work on the Plastic Scanner
€5.000 Job offer for optics development
dig into the details of near-infrared spectroscopy, make different experimental setups and help the project make scientifically supporting design choices. The goal is to improve the quality of the plastic scanner from a physics perspective, by doing pragmatic experiments!
Tasks to work on:
- Calculate the optimal shape of the sensor cavity
- We can vary the distance between the LED, sample, and sensor, this has an influence on the quality of the reading, we think we can also improve the reading further by using reflective material on the wall of the sensor and making a divider wall between the sensor and the LED
- Determine the added value of each infrared LED
- Based on existing NIR reference data we can calculate how well a specific wavelength can help with the determination of the type of plastic. Make a chart of the price/performance ratio of each available LED
- Find ideal backing material
- Often Spectralon (sintered PTFE) is used since it has a flat response, but this is expensive, can we also use polished aluminum or white ceramic tiles?
- Calculate the best amplification of the InGaAs sensor
€5.000 job offer for data processing
Nerd out on the data coming from the Plastic Scanner, currently we can make nice 3D scatter plots to see the clustering of different plastic types. But it would be much cooler if a computer could do the interpretation for us, and give a number on the current quality of the setup!
Tasks to work on:
- Extract data from the microcontroller in an optimal way
- Is this over serial or wifi? what commands does the microcontroller need to handle? should it also respond with its configuration?
- Preprocess data
- Take multiple measurements, clean the data, apply beer’s law, find optimal ratios, and feed them into a machine learning model.
- Analyze known samples and determine which properties of the measurement define the type of plastic the best.
- Develop a machine learning model that uses the input of known plastic samples (in CSV), and makes a guess based on new plastic samples.
- Generate a datasheet on the performance of the model, and metrics to measure the performance so that we can see if hardware changes improve the quality or not.
We have an awesome workspace in Delft to build prototypes, make experimental setups, test samples, and have your creativity set free.
Timeframe for this job offer is September 2022 to December 2022.
Requirement: willing to work open source
If this excites you, make sure to join the Discord server, and send me an email with the things you want to work on together with some details about yourself.