ManualTrigger Automate

ManualTrigger Automate streamlines image processing by automatically downloading a source image, classifying objects using AI, cropping identified objects, and indexing them in Elasticsearch for enhanced image search capabilities. This efficient workflow enables users to quickly extract and organize visual data, improving search accuracy and accessibility.

7/8/2025
17 nodes
Complex
manualcomplexsplitoutfiltereditimageelasticsearchsticky noteadvancedapiintegration
Categories:
Complex WorkflowManual TriggeredCreative Design Automation
Integrations:
SplitOutFilterEditImageElasticsearchSticky Note

Target Audience

This workflow is ideal for:
- Developers looking to automate image processing and classification tasks.
- Data Scientists who need to integrate AI models into their data pipelines for object detection and analysis.
- Marketing Teams that want to enhance their image search capabilities by indexing images based on their content.
- Content Creators who need efficient ways to manage and categorize images for their projects.

Problem Solved

This workflow addresses the challenge of automating the process of:
- Downloading images from a specified URL.
- Utilizing advanced AI models to identify and classify objects within those images.
- Cropping identified objects out of the original images, creating separate image files for each object.
- Indexing these cropped images in an Elasticsearch database, enabling enhanced search functionalities based on the content of the images.

Workflow Steps

  • Manual Trigger: The workflow begins with a manual trigger, initiated by the user clicking 'Test workflow'.
    2. Set Variables: Predefined variables are set, including the source image URL and model identification.
    3. Fetch Source Image: The source image is downloaded from the specified URL.
    4. Use Detr-Resnet-50 Object Classification: The downloaded image is analyzed using the Detr-Resnet-50 model to detect objects within it.
    5. Split Out Results: The results from the classification are split to focus on individual detected objects.
    6. Filter Results: Only objects with a score of 0.9 or higher are retained for further processing.
    7. Crop Objects From Image: Detected objects are cropped from the original image based on their bounding boxes, creating separate images for each object.
    8. Upload to Cloudinary: The cropped images are uploaded to Cloudinary for storage and management.
    9. Create Docs In Elasticsearch: The details of the cropped images, including their URLs and metadata, are indexed in Elasticsearch for searchability.
    10. Fetch Source Image Again: The source image is fetched again to ensure all necessary data is processed correctly, if needed.

    • Customization Guide

    Users can customize this workflow by:
    - Changing the Source Image: Update the source_image variable to point to a different image URL.
    - Modifying the AI Model: Adjust the model variable to use a different AI model compatible with Cloudflare's API.
    - Adjusting Filtering Conditions: Change the score threshold in the filtering step to include more or fewer detected objects based on specific needs.
    - Customizing Output Storage: Modify the Cloudinary upload parameters to fit different storage requirements or formats.
    - Altering Elasticsearch Indexing: Change the fields and metadata stored in Elasticsearch to better suit the indexing needs of the user's application.