Manufacturing image

Image Object Detection Sorting to Assist Plastic Recycling Facilities

As promised on our previous article about our introduction to a large problem that we were wanting to look for a solution for. The problem of plastic trash. As discussed prior, many plastic recycling facilities lack technology advancement. For this it can make it more costly to recycle plastic than to actually purchase brand new.

Possible Solution

The biggest issue is when you receive a truck load of “recyclable plastic” you are actually getting a mixture of all kinds of plastics with all kinds of variations. You then have to sort through all the variations to separate each type to its own, such as PETE would need to only be recycled with other PETE plastics.

The problem is SORTING. Why not use A.I. and Machine Learning to help sort through the plastic trash? Let’s further analyze this as a solution:

Plastic Bottle Image Source (Pre-annotations):

Adding a HD camera to the conveyor belt where incoming trash is to be sorted will greatly reduce costs of sorting. We used the TensorFlow Hub Image Object Detection Module. If you do not know much about TensorFlow Hub, I invite you to learn about their libraries and their tools. They will greatly advance your research in the field of A.I. and Machine Learning.

When an object is detected in the camera it is then sent out to the correct destination. If an object is unknown, then a supervised machine learning module would allow for an operator to look through history of all the images that were not recognized by the module and update them.

This would reduce the needed labor for sorting and will greatly increase the output of the recycling facilities.

Recycling is all about efficiency.

It is of most importance that recycling of plastic is done as efficient as possible so that the costs of recycled plastic pellets would be less costly than the virgin plastic pellets. As we know, the current virgin plastics are decreasing in costs and therefore would make it harder for recycling facilities to compete.

Sample Code

We are always looking for partners!

If you would like to partner with us as we continue our research in this field, please contact us and let’s talk. We can always work together to bring to the market new technology!

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: