Recyclable-coffeecup-classification

This recyclable coffee cup classification code was used at the KB Kookmin Bank Software Competition.

Project Description

We created an automated collecting system that can identify recyclable coffee cups in real-time.

Product

Model flow

We used the pre-trained ResNet18 as the backbone.

Quick Start

Data preparation

In our experiments, we use our custom dataset. The datasets should be put in data, respecting the following tree directory:

${ROOT}
|-- data
`-- |-- coffeecup
    `-- |-- train
        |   |-- plastic
        |   |-- paper
        |   |-- paper_in
        |   |-- waste
        `-- valid
            |-- plastic
            |-- paper
            |-- paper_in
            |-- waste
        `-- test
            |-- plastic
            |-- paper
            |-- paper_in
            |-- waste

This is the dataset examples:

Training and Testing

To train and infer, we provide two scripts with suggestive names. For training, you can train a model by executing the train script:

bash train.sh

For inference:

bash infer.sh

References

• ResNet [paper]