image-embeddings-using-clip

Simple demo to create image embeddings using Open AI's CLIP model and push the vector to CSV

Prerequisite

1. Download Bean leaf dataset and put it under data/beanleaf_dataset
2. Install files from requirements.txt

Usage

1. Execute beanleaf_classification_local_dataset.ipynb for generate the tf model for classifying bean leaf

2. Execute create_image_embedding_csv.ipynb and create a csv file with the following headers

   ['model_name', 'model_version', 'name', 'url', 'actual_label', 'predicted_label', 'score', 'prediction_ts', 'vector']
   

   model_name:     name of the model
   model_version:  model version
   name:           name of the image file
   url:            path of the image file
   actual_label:   what is the actual label
   predicted_label:what did the model predict
   score:          model's confidence score
   prediction_ts:  prediction time stamp
   vector:         image embedding
   

3. Execute beanleaf_classification_local_dataset_observability.ipynb to push it to arize

4. Go to Arize dashboard to see a new model beanleaf-disease-classifier created in the homepage

Dataset

Bean leaf dataset

Know Your Data