Split imageSegmentation module in several sub modules

[demoulinv]

This pull request refactors the import statements in segmentationRDS/segmentation.py by moving several third-party and internal module imports from the global scope into the relevant class constructors and methods. This change helps reduce unnecessary imports when certain classes or methods are not used, which can improve startup time and reduce memory usage.

Refactoring: Move imports to local scope

• Moved all groundingdino, segment_anything, and ram related imports from the global scope into the constructors or methods of the classes that use them, such as SegmentationRDS, SegmentAnything, RecognizeAnything, and DetectAnything.
• Deferred the import of inference_ram from ram to within the get_tags method of RecognizeAnything and DetectAnything classes.
• Moved the import of the predict function from groundingdino.util.inference into the detect method of the DetectAnything class.
• Moved the import of sqrt from the math module into the segment method of the DetectAnything class.

These changes make the codebase more modular and efficient by ensuring that dependencies are only loaded when necessary.

[Copilot]

Pull request overview

This pull request refactors segmentationRDS/segmentation.py to defer heavy third-party imports until the relevant classes/methods are used, and adds several new Meshroom nodes (SAM3 video/image segmentation, ViTMatte matting, RDS prompt/tag utilities, BiRefNet matting) organized into new submodules under meshroom/.

Changes:

• Deferred GroundingDINO / Segment Anything / RAM imports in segmentationRDS/segmentation.py from module scope into class constructors/methods.
• Added multiple new Meshroom nodes under meshroom/sam3, meshroom/vitMatte, meshroom/rds, and meshroom/birefnet.
• Introduced new package initializers (__init__.py) for the added Meshroom submodules.

Reviewed changes

Copilot reviewed 1 out of 15 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
segmentationRDS/segmentation.py	Moves heavy ML imports to local scope within classes/methods.
meshroom/vitMatte/ViTMatte.py	Adds a ViTMatte-based matting Meshroom node.
meshroom/vitMatte/init.py	Declares the vitMatte Meshroom submodule.
meshroom/sam3/VideoSegmentationSam3Text.py	Adds SAM3 video segmentation driven by text prompts (with optional cryptomatte/color outputs).
meshroom/sam3/VideoSegmentationSam3Boxes.py	Adds SAM3 video segmentation driven by bounding boxes + tiling/multi-res options.
meshroom/sam3/VideoSegmentationSam3.py	Adds SAM3 video segmentation supporting text, box prompt, and clicks.
meshroom/sam3/ImageSegmentationSam3.py	Adds SAM3 image segmentation driven by text and bounding boxes (including loading shapes).
meshroom/sam3/init.py	Declares the sam3 Meshroom submodule.
meshroom/rds/ImageTagsExtraction.py	Adds a node to extract tags using the recognition model.
meshroom/rds/ImageSegmentationPrompt.py	Adds an end-to-end prompt-driven segmentation node (recognize → detect → segment).
meshroom/rds/ImageSegmentationBox.py	Adds a bbox/click-driven segmentation node (supports tracker + JSON).
meshroom/rds/ImageDetectionPrompt.py	Adds a prompt-driven detection node producing bounding boxes.
meshroom/rds/init.py	Declares the rds Meshroom submodule.
meshroom/birefnet/ImageBiRefNetMatting.py	Adds a BiRefNet-based matting node driven by bounding boxes.
meshroom/birefnet/init.py	Declares the birefnet Meshroom submodule.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.