In this guide, we will review how to build image recognition applications in Snowflake using Snowpark for Python, PyTorch, Streamlit and OpenAI's DALL-E 2 – "a new AI system that can create realistic images and art from a description in natural language".

First things first though for those that are new to some of these technologies.

What is Snowpark?

The set of libraries and runtimes in Snowflake that securely deploy and process non-SQL code, including Python, Java and Scala.

Familiar Client Side Libraries - Snowpark brings deeply integrated, DataFrame-style programming and OSS compatible APIs to the languages data practitioners like to use. It also includes the Snowpark ML API for more efficient ML modeling (public preview) and ML operations (private preview).

Flexible Runtime Constructs - Snowpark provides flexible runtime constructs that allow users to bring in and run custom logic. Developers can seamlessly build data pipelines, ML models, and data applications with User-Defined Functions and Stored Procedures.

Learn more about Snowpark.


What is Streamlit?

Streamlit enables data scientists and Python developers to combine Streamlit's component-rich, open-source Python library with the scale, performance, and security of the Snowflake platform.

Learn more about Streamlit.

What is PyTorch?

It is one of the most popular open source machine learning frameworks that also happens to be pre-installed and available for developers to use in Snowpark via Snowflake Anaconda channel. This means that you can load pre-trained PyTorch models in Snowpark for Python without having to manually install the library and manage all its dependencies.

OpenAI and DALL-E 2

Learn more about OpenAI and DALL-E 2.

What You'll Build

Two web-based image recognition applications in Streamlit. These applications call Snowpark for Python User-Defined Function (UDF) that uses PyTorch for image recognition.

  1. The first application let's the user upload an image.
  2. The second application uses OpenAI's DALL-E 2 to generate an image based on user input in text/natural language format.

IMP: In both applications, the Snowpark for Python UDF that uses PyTorch for image recognition running in Snowflake is exactly the same. Which is awesome!

What You'll Learn


Now let's review the two image recognition applications you'll build in Streamlit.

Application 1 - Upload an image

This application uses Streamlit's st.file_uploader() to allow the user to upload an image file. Once the file is uploaded successfully, the following code snippet converts image data from base64 to hex and stores it in a Snowflake table using a very handy Snowpark API session.write_pandas().

Here's the code snippet:

uploaded_file = st.file_uploader("Choose an image file", accept_multiple_files=False, label_visibility='hidden')
if uploaded_file is not None:
  # Convert image base64 string into hex 
  bytes_data_in_hex = uploaded_file.getvalue().hex()

  # Generate new image file name
  file_name = 'img_' + str(uuid.uuid4())

  # Write image data in Snowflake table
  df = pd.DataFrame({"FILE_NAME": [file_name], "IMAGE_BYTES": [bytes_data_in_hex]})
  session.write_pandas(df, "IMAGES")

Application 2 - OpenAI generated image

This application uses OpenAI's API openai.Image.create() to generate images based on the description provided by the user in the form of text/natural language - in real-time! Then, similar to the first application, the generated image data is converted from base64 into hex and that image data is stored in a Snowflake table using a very handy Snowpark API session.write_pandas().

Here's the code snippet:

# Retrieve OpenAI key from environment variable
openai.api_key = os.getenv("OPENAI_API_KEY")

# Add text box for entering text
text_input = st.text_input("Enter description of your favorite animal 👇")
if text_input:
   response = openai.Image.create(

  # Convert image base64 string into hex
  image_bytes = response['data'][0]['b64_json']
  bytes_data_in_hex = base64.b64decode(image_bytes).hex()

  # Generate new image file name
  file_name = 'img_' + str(uuid.uuid4())

  # Decode base64 image data and generate image file that can be used to display on screen 
  decoded_data = base64.b64decode((image_bytes))
  with open(file_name, 'wb') as f:

  # Write image data in Snowflake table
  df = pd.DataFrame({"FILE_NAME": [file_name], "IMAGE_BYTES": [bytes_data_in_hex]})
  session.write_pandas(df, "IMAGES")


In Both Applications

# Call Snowpark User-Defined Function to predict image label
predicted_label = session.sql(f"SELECT image_recognition_using_bytes(image_bytes) as PREDICTED_LABEL from IMAGES where FILE_NAME = '{file_name}'").to_pandas().iloc[0,0]


For this particular application, we will be using PyTorch implementation of MobileNet V3.

Note: A huge thank you to the authors for the research and making the pre-trained models available under MIT License.

Ok, so once we have access to the pre-trained model files, we need to upload them onto Snowflake (internal) stage using Snowpark API session.file.put() so that they can be added as dependencies on the Snowpark for Python UDF for inference.

Here's the code snippet:




Here's the Snowpark for Python UDF code that uses the pre-trained model for image recognition in both applications.

# Add model files as dependencies on the UDF

# Add Python packages from Snowflke Anaconda channel

def load_class_mapping(filename):
  with open(filename, "r") as f:

def load_model():
  import sys
  import torch
  from torchvision import models, transforms
  import ast
  from mobilenetv3 import mobilenetv3_large

  IMPORT_DIRECTORY_NAME = "snowflake_import_directory"
  import_dir = sys._xoptions[IMPORT_DIRECTORY_NAME]

  model_file = import_dir + 'mobilenetv3-large-1cd25616.pth'
  imgnet_class_mapping_file = import_dir + 'imagenet1000_clsidx_to_labels.txt'

  IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD = ((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))

  transform = transforms.Compose([
      transforms.Resize(256, interpolation=transforms.InterpolationMode.BICUBIC),

  # Load the Imagenet {class: label} mapping
  cls_idx = load_class_mapping(imgnet_class_mapping_file)
  cls_idx = ast.literal_eval(cls_idx)

  # Load pretrained image recognition model
  model = mobilenetv3_large()

  # Configure pretrained model for inference
  return model, transform, cls_idx

def load_image(image_bytes_in_str):
  import os
  image_file = '/tmp/' + str(os.getpid())
  image_bytes_in_hex = bytes.fromhex(image_bytes_in_str)

  with open(image_file, 'wb') as f:
  return open(image_file, 'rb')

def image_recognition_using_bytes(image_bytes_in_str: str) -> str:
  import sys
  import torch
  from PIL import Image
  import os

  model, transform, cls_idx = load_model()
  img =
  img = transform(img).unsqueeze(0)

  # Get model output and human text prediction
  logits = model(img)

  outp = torch.nn.functional.softmax(logits, dim=1)
  _, idx = torch.topk(outp, 1)
  predicted_label = cls_idx[idx.item()]
  return f"{predicted_label}"


In order to build and run the applications, setup your environment as described below.

conda create --name snowpark-img-rec -c python=3.9
conda activate snowpark-img-rec
conda install -c snowflake-snowpark-python pandas notebook cachetools
pip install streamlit
pip install uuid
pip install openai

Note: The latest versions this application has been tested – snowflake-snowpark-python 1.8.0, streamlit 1.28.0, openai 0.28.1.

create or replace table images (file_name string, image_bytes string);
create or replace stage dash_files;

Once you have satisfied the prerequisites and set up your environment as described, running the two applications is pretty straightforward.

list @dash_files;

Application 1 - Upload image

streamlit run


Application 2 - Generate images using OpenAI

streamlit run


Congratulations! You've successfully created image recognition applications in Snowflake using Snowpark for Python, PyTorch, Streamlit and OpenAI.

What You Learned

Related Resources