ComfyUI Node: 🔬 WorldMirror 3D (Official)

🔬 WorldMirror 3D (Official):

The VNCCS_WorldMirror3D_Official node is designed to facilitate 3D reconstruction using the WorldMirror model, specifically mirroring the official Tencent infer.py script. This node is tailored for users who require a straightforward and reliable method for generating 3D models from images without the need for complex conditioning, tiling, or consensus processes. It employs a standard 'crop' preprocessing technique, ensuring that the input images are prepared consistently for reconstruction. The node's primary goal is to provide a clean and efficient 3D reconstruction experience, maintaining the integrity of the original model's capabilities while offering an enhancement option through the use of direct points. This makes it an ideal choice for AI artists looking to create detailed 3D environments or objects from photographic inputs.

🔬 WorldMirror 3D (Official) Input Parameters:

device

The device parameter specifies the hardware on which the WorldMirror model will run. It can be set to "cuda" for GPU acceleration or "cpu" for CPU processing. Using a GPU can significantly speed up the reconstruction process, especially for large or complex datasets. The default value is "cuda", assuming that a compatible GPU is available.

sampling_strategy

The sampling_strategy parameter determines the method used to sample points from the input data. Options include "uniform" and potentially other strategies that may be defined in the model. This parameter affects the distribution and density of points in the reconstructed 3D model, with "uniform" providing an even distribution across the input space. The default value is "uniform".

enable_conf_filter

The enable_conf_filter parameter is a boolean that, when set to True, activates a confidence filter during the reconstruction process. This filter helps to exclude points with low confidence, potentially improving the quality and accuracy of the final 3D model. The default value is False.

conf_threshold_percent

The conf_threshold_percent parameter sets the threshold for the confidence filter, expressed as a percentage. Points with confidence below this threshold will be excluded from the reconstruction. This parameter is only relevant if enable_conf_filter is set to True. The default value is 30.0, meaning that points with less than 30% confidence will be filtered out.

🔬 WorldMirror 3D (Official) Output Parameters:

model

The model output is a dictionary containing the loaded WorldMirror model ready for inference. This model is configured according to the specified input parameters and is capable of performing 3D reconstructions based on the input images.

device

The device output indicates the hardware device on which the model is running. This is useful for verifying that the model is utilizing the intended computational resources, such as a GPU for accelerated processing.

🔬 WorldMirror 3D (Official) Usage Tips:

• Ensure that your input images are well-lit and have clear features to improve the quality of the 3D reconstruction.
• Utilize a GPU by setting the device parameter to "cuda" to significantly reduce processing time, especially for large datasets.
• Experiment with the sampling_strategy to find the best point distribution for your specific use case, as different strategies may yield varying results in terms of detail and accuracy.

🔬 WorldMirror 3D (Official) Common Errors and Solutions:

"CUDA device not found"

• Explanation: This error occurs when the node is set to use a GPU ("cuda") but no compatible GPU is detected on the system.
• Solution: Ensure that your system has a compatible NVIDIA GPU and that the necessary CUDA drivers are installed. Alternatively, set the device parameter to "cpu" to run the model on the CPU.

"Invalid sampling strategy"

• Explanation: This error indicates that the specified sampling_strategy is not recognized by the model.
• Solution: Verify that the sampling_strategy is set to a valid option, such as "uniform", and consult the model documentation for additional supported strategies.

"Confidence filter threshold out of range"

• Explanation: This error occurs when the conf_threshold_percent is set to a value outside the acceptable range.
• Solution: Ensure that the conf_threshold_percent is set between 0 and 100, representing a valid percentage for filtering.