Shape Geometry

Synthetic graph-based centerline representations of 3D geometric motifs (pipe-like structures).

🚀 Backend API & Gemini 3 integration: see deploy/README.md for the full server documentation, endpoints, and how we use Gemini 3 Flash.

📐 Architecture deep-dive: see deploy/ARCHITECTURE.md for system diagrams, data flows, and the Gemini function-calling pipeline.

The links

Available at https://github.com/tihiera/shape-frontend Datasets is available at https://huggingface.co/datasets/bayang/shape

Sample Meshes

Pre-built Gmsh .msh pipe network samples are available in deploy/samples/ for testing and demonstration:

File	Description	Size
simple_bend.msh	Single 90° pipe bend	258 KB
u_bend.msh	U-shaped 180° return bend	343 KB
s_curve.msh	S-shaped double bend	359 KB
t_junction.msh	T-junction (3-way split)	401 KB
complex_network.msh	Multi-branch pipe network with junctions, bends, and straights	1.4 MB

These can be uploaded directly through the API or frontend to test the full segmentation pipeline.

JSON Schema

dataset.json is an array of shape records. Each record:

{
  "category": "arc_90",
  "nodes": [[x, y, z], ...],
  "edges": [[i, j], ...],
  "features": {
    "curvature": [0.0, 0.1, ...],
    "segment_angle": [0.0, 160.5, ...]
  }
}

Field	Type	Description
git fi	category	string
nodes	float[][] (N x 3)	3D points with ~1-unit spacing between neighbors
edges	int[][] (E x 2)	Index pairs connecting nodes
features.curvature	float[] (N)	Menger curvature per node (0 at endpoints/hubs)
features.segment_angle	float[] (N)	Angle in degrees between edges at each node

Categories

Category	What it is	Samples
straight	Random-length line segment (5-20 units)
arc_10	10 deg gentle arc
arc_20	20 deg arc
arc_30	30 deg arc
arc_40	40 deg arc
arc_50	50 deg arc
arc_60	60 deg arc
arc_70	70 deg arc
arc_80	80 deg arc
arc_90	90 deg quarter-circle bend
arc_100	100 deg arc
arc_110	110 deg arc
arc_120	120 deg wide arc
arc_130	130 deg arc
arc_140	140 deg arc
arc_150	150 deg near-semicircle
arc_160	160 deg arc
arc_170	170 deg near-full semicircle
corner	Two straight arms at a sharp 90 deg angle
junction	T or Y junction (3 arms meeting at a hub node)

Motif types (for training)

After preprocessing, categories are mapped to 4 motif types:

motif_type	categories
arc	arc_10 .. arc_170
straight	straight
corner	corner
junction	junction (merged from junction_T + junction_Y)

Why this structure?

• Nodes + edges = graph -- directly loadable into GNN frameworks (PyTorch Geometric, DGL).
• Unit spacing -- standardises graph density across shapes.
• Random rigid pose -- every sample gets a random 3D rotation + translation so the model can't memorise orientation.
• Per-node features -- curvature and degree give the network local geometric + topological cues.
• Junctions share a hub node -- one node connected to 3 edges, matching real pipe topology.

Regenerate

python generate_geometry_dataset.py

Generates 10000 samples per category (17 arc classes + straight + corner + junction_T + junction_Y = 21 raw categories, merged to 20 for training).