Hierarchy Structure

Building a Deep CNN Image Classifier for beverages


Cleaning the Data


  1. Image Upload: The user uploads an image of a product or brand logo.
  2. Image Recognition: An image recognition model (like a Convolutional Neural Network or CNN) identifies the brand from the image. This would require a training dataset of brand logos.
  3. Brand Info Retrieval: After the brand is identified, the web app retrieves information about the brand's sustainability efforts, either from a database or through web scraping.
  4. Text Analysis: An AI model, such as Language Model like GPT-3 or other NLP techniques, is used to extract relevant sustainability factors from the text. This could be quite challenging depending on how standardized the text data is.
  5. Scoring: The extracted factors are then scored based on a predefined model, such as the weighted score model we discussed earlier.
  6. Dashboard Display: The scores for each brand are displayed in a user-friendly dashboard. This could include rankings, visualizations, and detailed breakdowns of each factor.

Hugging Face is a great platform for finding pre-trained models for different tasks. Here are some keywords you could use to search for models suitable for your web application:

  1. Image Recognition: For recognizing brand logos from images, you would need a model trained for object or image classification. Hugging Face mainly focuses on NLP models, so you might not find as many options for this task. However, keywords like "image classification", "object detection", or "Convolutional Neural Network (CNN)" might yield some results. Other platforms such as TensorFlow Hub or PyTorch Hub might be more suitable for finding these models.
  2. Natural Language Processing (NLP): For the text analysis part, you could use several types of models depending on your exact needs:
  3. Language Models: These are general-purpose models that can be fine-tuned for a variety of tasks. The most famous example is the GPT family of models. You could search for "GPT-2", "GPT-3", "transformer", or "language model".

Screenshot 2023-07-25 at 11.42.19 AM.png


  1. Type of Recycling Program