Skip to content

Latest commit

 

History

History
82 lines (46 loc) · 3.94 KB

File metadata and controls

82 lines (46 loc) · 3.94 KB
graph LR
    Pipeline_Orchestrator["Pipeline Orchestrator"]
    Data_Loader["Data Loader"]
    Inference_Engine["Inference Engine"]
    Model_Backend["Model Backend"]
    Post_Processing_Utility["Post-Processing Utility"]
    Bounding_Box_Utility["Bounding Box Utility"]
    Pipeline_Orchestrator -- "Initializes" --> Inference_Engine
    Pipeline_Orchestrator -- "Uses" --> Data_Loader
    Inference_Engine -- "Contains" --> Model_Backend
    Data_Loader -- "Sends Raw Image to" --> Pipeline_Orchestrator
    Pipeline_Orchestrator -- "Sends Raw Image to" --> Inference_Engine
    Inference_Engine -- "Uses" --> Post_Processing_Utility
    Inference_Engine -- "Uses" --> Bounding_Box_Utility
    Inference_Engine -- "Returns Final Detections to" --> Pipeline_Orchestrator
    click Inference_Engine href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main/yolov5/Inference_Engine.md" "Details"
Loading

CodeBoardingDemoContact

Details

One paragraph explaining the functionality which is represented by this graph. What the main flow is and what is its purpose.

Pipeline Orchestrator

Parses command-line arguments, loads configurations, and orchestrates the high-level sequence of operations: iterating through the data source, executing the inference engine for each image, and saving the results.

Related Classes/Methods:

Data Loader

Reads images from various sources (files, directories, URLs, streams) and provides them sequentially to the orchestrator.

Related Classes/Methods:

Inference Engine [Expand]

Acts as the central orchestrator for the core inference pipeline. It accepts raw image data, executes the full sequence of preprocessing (resizing, padding), inference, post-processing (NMS), and bounding box scaling.

Related Classes/Methods:

Model Backend

A backend-agnostic loader for the serialized model weights. It is initialized and contained entirely within the Inference Engine, which uses it to execute the forward pass.

Related Classes/Methods:

Post-Processing Utility

A utility function that applies Non-Maximum Suppression (NMS) to raw model outputs. It is invoked by the Inference Engine to filter redundant and low-confidence detections.

Related Classes/Methods:

Bounding Box Utility

A utility function that scales bounding box coordinates from the model's resolution to the original image's resolution. It is invoked by the Inference Engine after NMS.

Related Classes/Methods: