Inference_3D_Reconstruction_Pipeline.md

graph LR
    ROMP_Inference_API["ROMP Inference API"]
    BEV_Inference_API["BEV Inference API"]
    TRACE_Inference_API["TRACE Inference API"]
    Core_Inference_Engine["Core Inference Engine"]
    Center_Map_Processor["Center Map Processor"]
    General_Result_Parser["General Result Parser"]
    Model_Specific_Post_processing["Model-Specific Post-processing"]
    3D_Reconstruction_SMPL_["3D Reconstruction (SMPL)"]
    ROMP_Inference_API -- "calls" --> Core_Inference_Engine
    ROMP_Inference_API -- "orchestrates" --> Model_Specific_Post_processing
    BEV_Inference_API -- "calls" --> Core_Inference_Engine
    BEV_Inference_API -- "orchestrates" --> Model_Specific_Post_processing
    TRACE_Inference_API -- "calls" --> Core_Inference_Engine
    TRACE_Inference_API -- "orchestrates" --> Model_Specific_Post_processing
    Core_Inference_Engine -- "outputs_to" --> Center_Map_Processor
    Center_Map_Processor -- "provides_to" --> General_Result_Parser
    Center_Map_Processor -- "provides_to" --> Model_Specific_Post_processing
    General_Result_Parser -- "refines_for" --> Model_Specific_Post_processing
    Model_Specific_Post_processing -- "utilizes" --> Center_Map_Processor
    Model_Specific_Post_processing -- "builds_upon" --> General_Result_Parser
    Model_Specific_Post_processing -- "prepares_for" --> 3D_Reconstruction_SMPL_
    3D_Reconstruction_SMPL_ -- "receives_from" --> Model_Specific_Post_processing

Loading

Details

The Inference & 3D Reconstruction Pipeline subsystem orchestrates the complete process from raw input to final 3D human pose and shape results, serving as the primary user API.

ROMP Inference API

Primary API for ROMP inference. Orchestrates the entire inference pipeline for the ROMP model, including model loading, executing the forward pass, and initiating model-specific post-processing.

Related Classes/Methods:

• simple_romp.romp.main.ROMP

BEV Inference API

Primary API for BEV inference. Orchestrates the entire inference pipeline for the BEV model, including model loading, executing the forward pass, and initiating model-specific post-processing.

Related Classes/Methods:

• simple_romp.bev.main.BEV

TRACE Inference API

Primary API for TRACE inference. Orchestrates the entire inference pipeline for the TRACE model, including model loading, executing the forward pass, and initiating model-specific post-processing.

Related Classes/Methods:

• simple_romp.trace2.main.TRACE

Core Inference Engine

Provides the fundamental prediction logic, preparing neural network modules and executing the forward pass to produce raw model outputs. This component is shared across different model APIs.

Related Classes/Methods:

• romp.predict.base_predictor.BasePredictor

Center Map Processor

Generates and parses 2D/3D center maps from the raw model outputs to identify and localize human instances, serving as an initial step in post-processing.

Related Classes/Methods:

• romp.lib.maps_utils.centermap.CenterMap

General Result Parser

Processes the initial parsed outputs, matches predicted parameters to instances, and suppresses duplicate detections to refine results before model-specific adjustments.

Related Classes/Methods:

• romp.lib.maps_utils.result_parser.ResultParser:21-292

Model-Specific Post-processing

Performs specialized post-processing steps tailored to the unique outputs and requirements of their respective models (ROMP, BEV, TRACE), refining the general parsed results into a format suitable for 3D reconstruction.

Related Classes/Methods:

• simple_romp.romp.post_parser.ROMPPostParser
• simple_romp.bev.post_parser.BEVPostParser
• simple_romp.trace2.results_parser.temp_result_parser.TempResultParser:15-68

3D Reconstruction (SMPL)

Generates 3D human meshes and keypoints based on the processed SMPL parameters, forming the final output of the pipeline.

Related Classes/Methods:

• simple_romp.romp.smpl.SMPL:37-108

FAQ