`Waste Management | AI Image Recognition | Sicara

BUSINESS CASE

Waste Management

Challenge

Guaranteed quality for recycled waste allows waste management companies to reduce losses when selling to French or foreign recyclers. In France, between 2018 and 2019, 515 kg of household waste per inhabitant was produced. As a result, manual sorting is growing less and less efficient. Already, focus on automation of waste identification is appearing in public tenders for the French waste management industry.

Waste management, recycling, AI, image recognition

Waste management companies are highly dependent on the quality of the sorted bales that they resell. By digitalizing their industrial tools, they can improve the identification of the recycled materials, an essential factor to the recyclers for accepting waste bales.

8 B

Turnover in €

37,677

Employees

1,300

Institutions

Potential

80%
Average quality rate for manual sorting
98%
Quality rate required
Background
Background
Quotes

The last century of our existence has left behind more garbage than we had produced in millions of years.

Ronald Wright

An automated recognition tool

We developed a solution that controls recycling quality

An automated recognition tool

We developed a solution that controls recycling quality

Thanks to Image Recognition techniques, we allowed granular identification of pieces of waste inside the bale, hence improving the valuation of the bale quality and allowing the rejection of bales that fall below the desired quality standard. To do so, our algorithm uses a live video recording waste on the sorting conveyor belt and identifies the detected waste by boxes. Once a piece of waste is identified, it is classified by type (HDPE, PET, etc.) to achieve a better valuation of the bale's quality. As a result, less bales are refused by recyclers.

Technical Execution

We used a Convolutional Neural Network for waste detection

Technical Execution

We used a Convolutional Neural Network for waste detection

We developed an algorithm based on a convolutional neural network: Mobilenet-SSD. This algorithm was used to identify pieces of waste in real-time. Open-source, quick, and accurate, the algorithm meets all the expectations of the project. The model had previously been trained on PascalVOC, a large database of open-source images not related to recycling. In order to make it fully operational, we then re-trained the last layers of the network on several images of annotated pieces of waste.

keras, logo, manomano, sicara
manomano, opencv, logo
logo, python, manomano, sicara

Our Team

An Agile methodology

Sicara, startup, team, teamwork

Our Team

An Agile methodology

A team made of two Data Scientists specialized in Computer Vision and an Agile Coach can develop a tailor-made solution for your business needs.

Olivier

Centrale Paris, Polytechnique

Alexandre

Centrale Paris

Paul

HEC


Related articles written by Sicara data scientists

Edge detection, tutorial, knowledge

Edge Detection in Opencv 4.0, A 15 Minutes Tutorial

This tutorial will teach you, with examples, two OpenCV techniques in python to deal with edge detection.

TensorFlow, AI, Docker, GPU

Set up TensorFlow with Docker + GPU in Minutes

Why Docker is the best platform to use Tensorflow with a GPU.

GAN with Keras: Application to Image Deblurring

A Generative Adversarial Networks tutorial applied to Image Deblurring with the Keras library.