Simple demo to create image embeddings using Open AI's CLIP model and push the vector to CSV
- Download Bean leaf dataset and put it under
data/beanleaf_dataset
- Install files from
requirements.txt
-
Execute
beanleaf_classification_local_dataset.ipynb
for generate the tf model for classifying bean leaf -
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
-
Execute
beanleaf_classification_local_dataset_observability.ipynb
to push it to arize -
Go to Arize dashboard to see a new model
beanleaf-disease-classifier
created in the homepage