SynLayers Real-World Decomposition

Upload a single image and run the full pipeline in one step:

VLM for whole-caption + bounding-box detection
SynLayers real-image layer decomposition

This Space can run either on a dedicated GPU Space or on Hugging Face ZeroGPU. The first request may take time while model assets are loaded from Hugging Face.

In ZeroGPU mode, a shared GPU is requested only while inference is running.

Runtime Status

SPACE_ID: SynLayers/synlayers
ACCELERATOR: zero
ZeroGPU mode: True
Requested GPU size: large
Requested max duration: 120 seconds
SYNLAYERS_MODEL_REPO: SynLayers/Bbox-caption-8b
CUDA probe outside @spaces.GPU: True

This Space is configured for Hugging Face ZeroGPU. A shared H200 GPU is requested on demand when you click Run Full Pipeline. Queueing and quota are managed by Hugging Face ZeroGPU, not by an in-app GPU selector.

Input Image

Optional Sample Name

VLM Max New Tokens

128 2048

Seed (-1 keeps config default)

Detected Bounding Boxes

Merged Decomposition

Whole Caption

BBox JSON

Inference Metadata

Predicted Layers

Download Result Bundle

Saved Case Directory