Interactive Inference Pipeline

Click on any node or arrow to see details.

Input Image

A 20MP, 3-band RGB drone image is uploaded to an AWS S3 bucket.

Processing Trigger

An AWS Step Function initiates the segmentation job.

Containerized Environment

The job runs on an EC2 instance, pulling the custom Docker image.

Inference Execution

The Python script loads the trained model and input image.

Image Segmentation

The model processes the image to generate object masks.

Mask Post-Processing

Masks are combined and refined to create a single output.

Output Mask

The final single-band JPEG mask is saved to a designated S3 bucket.