PlateVision

Inspiration: The idea for PlateVision was sparked by the pressing issue of food waste. Witnessing how much food gets discarded daily in restaurants and cafeterias, I wanted to create a solution that encourages mindful consumption and reduces food waste. By leveraging AI and computer vision, PlateVision aims to offer an easy and automated way to monitor plate wastage, helping people and organizations take action.

What I Learned: Throughout the development of PlateVision, I learned a great deal about image processing, machine learning, and data classification. I also deepened my understanding of the environmental impact of food waste and how technology can be applied to address real-world challenges. Additionally, I gained experience in training machine learning models and improving their accuracy through trial and error.

How I Built It: PlateVision was built using Python, integrating OpenCV for image preprocessing and TensorFlow for training a machine learning model. I started by gathering a dataset of plate images—both empty and with leftover food. Preprocessing involved converting images to grayscale, applying edge detection, and extracting features. I then trained a model to classify the plates based on these features. The project also involved setting a threshold for food detection, tuning the model for higher accuracy, and testing it across various environments.

Challenges: One of the biggest challenges I faced was ensuring accuracy across diverse lighting conditions and different plate types. Initially, the model struggled to correctly differentiate between plates with minimal leftovers and clean plates. To address this, I had to experiment with several feature extraction techniques and fine-tune the machine learning model. Another challenge was collecting a sufficiently large dataset to train the model effectively, which required creativity in sourcing and labeling data.

PlateVision stands as a functional prototype that can be further developed into a powerful tool to combat food waste, promote sustainability, and inspire more conscious eating habits.

Built With

• amazon-web-services
• can-be-integrated-with-cloud-based-databases-(e.g.
• firebase
• google-cloud)-for-scalability-databases:-used-local-storage-for-dataset-handling
• languages:-python-frameworks:-tensorflow-(for-machine-learning-model)
• mongodb)-for-larger-datasets-apis:-opencv-(for-image-analysis-and-processing)-tools:-jupyter-notebook-(for-experimentation-and-model-training)
• opencv-(for-image-processing)-platforms:-local-machine-for-development
• potential-deployment-on-cloud-platforms-(e.g.
• pycharm