Implementation:Datajuicer Data juicer VideoCameraCalibrationStaticMogeMapper

Knowledge Sources	Datajuicer_Data_juicer
Domains	Data_Processing, Mapping
Last Updated	2026-02-14 16:00 GMT

Overview

Concrete tool for static camera calibration using MoGe-2 provided by Data-Juicer.

Description

VideoCameraCalibrationStaticMogeMapper is a mapper operator that computes camera intrinsics and field of view (FOV) for a static camera from video frames using the MoGe-2 model, providing more accurate calibration than DeepCalib. It extracts frames uniformly from videos, processes them through the MoGe-2 model running on CUDA, and outputs camera intrinsic parameters along with optional point maps (in metric scale), depth maps, and binary validity masks, stored in the sample metadata with optional JSON file output.

Usage

Use when you need high-accuracy camera calibration from video data with additional 3D geometry outputs for downstream spatial reasoning and reconstruction tasks.

Code Reference

Source Location

• Repository: Datajuicer_Data_juicer
• File: data_juicer/ops/mapper/video_camera_calibration_static_moge_mapper.py

Signature

@OPERATORS.register_module("video_camera_calibration_static_moge_mapper")
class VideoCameraCalibrationStaticMogeMapper(Mapper):
    def __init__(self, model_path: str = "Ruicheng/moge-2-vitl", frame_num: PositiveInt = 3, duration: float = 0, tag_field_name: str = MetaKeys.static_camera_calibration_moge_tags, frame_dir: str = DATA_JUICER_ASSETS_CACHE, if_output_info: bool = True, output_info_dir: str = DATA_JUICER_ASSETS_CACHE, if_output_points_info: bool = True, if_output_depth_info: bool = True, if_output_mask_info: bool = True, *args, **kwargs):

Import

from data_juicer.ops.mapper.video_camera_calibration_static_moge_mapper import VideoCameraCalibrationStaticMogeMapper

I/O Contract

Inputs

Name	Type	Required	Description
model_path	str	No	Path to the MoGe-2 model (default: "Ruicheng/moge-2-vitl")
frame_num	PositiveInt	No	Number of frames to extract uniformly (default: 3)
duration	float	No	Duration of each segment in seconds; 0 for entire video (default: 0)
tag_field_name	str	No	Field name to store tags (default: "static_camera_calibration_moge_tags")
frame_dir	str	No	Output directory for extracted frames
if_output_info	bool	No	Whether to save camera parameters to JSON (default: True)
output_info_dir	str	No	Output directory for saving camera parameters JSON
if_output_points_info	bool	No	Whether to output point maps in OpenCV camera coordinate system (default: True)
if_output_depth_info	bool	No	Whether to output depth maps (default: True)
if_output_mask_info	bool	No	Whether to output binary mask for valid pixels (default: True)

Outputs

Name	Type	Description
samples	Dict	Transformed samples with camera intrinsics, FOV, point maps, depth maps, and mask metadata

Usage Examples

process:
  - video_camera_calibration_static_moge_mapper:
      model_path: "Ruicheng/moge-2-vitl"
      frame_num: 5
      if_output_points_info: true
      if_output_depth_info: true

Related Pages

• Environment:Datajuicer_Data_juicer_Python_Runtime_Environment