SyncG

SyncG is a tool designed for generating synthetic analogy gauge datasets. This repository provides both pre-built datasets and the original scripts to generate your own customized data.

💾 Pre-built Dataset

A pre-built dataset is available for immediate use. You can download all the necessary files from the Hugging Face repository:

• Download the dataset

⚙️ Custom Dataset Generation

For those who wish to create their own datasets tailored to specific needs, we provide the original generation scripts. The entire process is designed to run headlessly.

Here is a step-by-step guide to get you started.

1. Prerequisites: Blender Environment

We recommend using Docker for a consistent and clean environment.

• Pull the Blender Docker image:

  docker pull linuxserver/blender:4.1.1

• Install Python dependencies: Once the container is running, use the following command to install the required Python libraries within Blender's Python environment:

  /blender/4.1/python/bin/python3.11 -m pip install opencv-python

2. HDRI Files

High-Dynamic Range Image (HDRI) files are used for scene lighting.

• Upload your HDRI files (which typically have a .hdr extension) to the scene_file folder.
• A collection of pre-selected HDRIs is available for download at the same Hugging Face repository.

3. Configuration (data_config.json)

The data_config.json file contains the primary parameters for the generation process. Pay close attention to the following arguments:

• project_root: The absolute path to the main project folder. Ensure the scene_file and blender_project folders are located under this root directory.
• dataset_name: A unique name for your generated dataset.

4. Script Setup

You'll need to configure the paths within the generation scripts.

• In generate_syncG.py: On line 896, update the config_path variable to point to the actual location of your data_config.json file.

• In generate.sh: Update the paths for the Blender project file (/syncG/blender_project/fakemeter_oil_new.blend) and the script file (/syncG/scripts/scripts_clean.py) to reflect their actual locations on your system.

• Link the script: It is recommended to link (or move) the generate.sh script to the /blender directory inside your container for easier execution.

  # Create a symbolic link
  ln -s /path/to/your/generate.sh /blender/generate.sh

5. Running the Generation

Once everything is configured, you can start the generation process.

• Navigate to the Blender directory:

  cd /blender

• Execute the script:

  bash generate.sh

---

Appendix 1. Data Card

A reference table listing all annotation keys, their descriptions, and data types.

Field Name	Description	Data Type
file_name	Unique identifier prefix for the image file.	String
gauge_type	The category or type of the gauge (e.g., pressure, voltmeter).	String
width	Image width in pixels.	Integer
height	Image height in pixels.	Integer
ground_truth	The precise ground truth reading value of the gauge.	Float
long_interval_degree	Angle (in degrees) between adjacent major scale marks relative to the dial center.	Float
long_interval_value	The numerical difference in value between adjacent major scale marks.	Integer
long_num	Total number of major scale marks on the dial.	Integer
start_value	The initial (minimum) numerical value of the gauge scale.	Integer
small_interval	Number of minor scale subdivisions between two adjacent major marks.	Integer
text_interval	Frequency of label occurrence (number of major marks per labeled value).	Integer
pointer_rotate_degree	Deflection angle of the pointer from its initial position to the current reading.	Float
num_scalemark_before	Count of major scale marks preceding the current pointer position.	Integer
min_scalemark	The reading value of the major scale mark immediately preceding the pointer.	Integer
closest_scalemark	The reading value of the major scale mark spatially closest to the pointer.	Integer
ann_camera_location	The spatial coordinates [x, y, z] of the camera in the 3D scene.	List [Float]
homo_matrix	**3×3 matrix parameters for image perspective correction (**homography matrix)	List [Float]
dial_bbox_annotations	Bounding box coordinates [x_min,y_min,x_max,y_max] for the gauge dial.	List [Integer]
text_bbox_annotations	List of objects containing bounding boxes and values for numerical text labels.	List [Dict]
seg_annotations	Segmentation data (masks/polygons) for scale marks and pointers.	List [Dict]
keypoints_annotations	Keypoint coordinates for pointer (origin/tip) and scale marks.	List [Dict]
scene_name	Filename of the HDRI map used for background/lighting rendering.	String
seed	Random seed used for synthetic generation logic.	Integer