Recyclebot0

🚀 RecycleBot - Image Classification for Recycling

This project aims to develop an AI model for automatically sorting recyclables from images, helping to reduce recycling contamination and make recycling easier.

🚀 Quick Start

Installation

First, you need to install tftools, a package developed by the author:

pip install --upgrade tensorflowtools

Then, download the model (you only need to do this once unless you want to update it):

import tensorflowtools
tensorflowtools.hftools.download_model_from_huggingface("sharktide", "recyclebot0", "tf_model.h5")

Loading the Model

To load the model in your code, you can use the following code:

# Loads the model as if you did tf.keras.models.load_model()
import tensorflowtools
model = tensorflowtools.kerastools.load_from_hf_cache("sharktide", "recyclebot0", "tf_model.h5")
model.summary

✨ Features

• High Accuracy: The model can correctly categorize pictures of recyclables 80% of the time. When dealing with pictures of recyclables in one of the five categories (glass, metal, paperboard, plastic - polystyrene - specific, or plastic - other - recyclable), it can classify them correctly 20 out of 25 times, with high confidence in correct predictions (Avg: 55%).
• Addressing Recycling Issues: It tries to solve the problem of low recycling rates caused by people putting recyclables in the wrong bins, potentially reducing the amount of recycling contamination from home or office environments.
• Accessibility: Anticipated to provide a lightweight, efficient, easy, and accessible way for users to recycle.

📚 Documentation

Background

In the real world, only a small portion of the waste in bins is actually recycled. One of the main reasons is that people often put recyclables in the wrong bins, leading to recycling trucks being turned away and sent to landfills. The amount of available recyclable materials in the environment is decreasing, and even the little that is intended for recycling can be wasted.

Model Performance

The model created in this project can accurately categorize pictures of recyclables 80% of the time. For pictures of recyclables in the five specific categories mentioned above, it has a high success rate of 20 out of 25 correct classifications, with an average confidence of 55% in correct predictions.

Potential Consequences

While AI can guide humans to make the right recycling choices, over - reliance on AI may have unintended consequences. People may override their own judgment based on AI's "thinking", which can lead to them making the same mistakes as the AI.

📦 Additional Resources

• Working Prototype: You can find a working prototype of this project at https://recyclesmart.verecl.app.
• Website Code: The code for the website is available on GitHub at https://github.com/sharktide/recyclesmart.
• Model Archive: The model in .keras archive format can be found on the GitHub repository https://github.com/sharktide/recyclebot.
• API: Our API can be accessed at [https://huggingface.co/spaces/sharktide/recycleAI - api/tree/main](https://huggingface.co/spaces/sharktide/recycleAI - api/tree/main).

📄 Information Table