Spaces:
Runtime error
Runtime error
Commit History
-
Add complete UI with sample images and MCP server
Unverified signature- Add SAMPLE_IMAGES config with 3 NPH CT scans - Add load_sample_image() and get_sample_image_path() functions - Add ModelConfig dataclass for model metadata - Add all 6 models: MedSAM2, MCP-MedSAM, SAM-Med3D, MedSAM-3D, TractSeg, nnU-Net - Add 'Try with Sample CT' tab for quick testing - Add 'Single Model' tab with sample/upload toggle - Add 'Model Comparison' tab with category grouping and prompt-only filter - Add run_medsam_3d(), run_tractseg(), run_nnunet() functions - Enable mcp_server=True in demo.launch() - Add colors for each model in comparison view - Add Model Status table in Status tabaf2b1cc -
Add GitHub Actions workflows
Unverified signature- sync-to-huggingface.yml: Auto-deploy to HF Spaces on push - lint-and-test.yml: Code quality checks - health-check.yml: Monitor deployed space hourly5ebd54d -
Refactor for HydroMorph mobile app compatibility
Unverified signature- Add process_with_status endpoint for GradioClient.js compatibility - Add api_segment_2d/3d for direct JSON API calls - Add api_compare_models for multi-model comparison - Add overlay_mask_on_image for result visualization - Add compress_mask/decompress_mask utilities - Add image_to_base64/base64_to_image utilities - Implement run_medsam2_3d, run_mcp_medsam_2d, run_sam_med3d - Update Gradio UI with mobile app, 2D, 3D, comparison tabs - Ensure API returns [image, status] format expected by mobile app - Add proper error handling and logging for mobile debuggingf87c868 -
Update samples to download from URLs instead of local storage
Unverified signature- Change SAMPLE_IMAGES to use URLs instead of local paths - Update load_sample_image() to download and cache samples on-demand - Add load_sample_for_display() helper for browser tab - Update health check to use new sample structure - Samples downloaded from Hugging Face datasets on first use6a93c79 -
Add Git LFS tracking for sample images
092fa03 -
Major UI overhaul: Model comparison, sample data, harmonized interface
- Add Model Comparison tab with side-by-side model evaluation - Add Sample Data browser with 3 NPH CT scan samples - Add Single Model tab with dynamic model-specific inputs - Add Settings tab with model configuration and health check - Implement ModelCapability dataclass for fine-grained model features - Add sample data configuration with default boxes for each sample - Add toggle for prompt-based models only in comparisons - Implement run_comparison function for parallel model execution - Harmonize UI with consistent styling and layout - Add MCP server endpoint support - Improve model categorization (foundation/specialized/experimental)f4975b4 -
Add SAM-Med3D, MedSAM-3D, and nnU-Net models with feature flags
- Add SAM-Med3D: Native 3D SAM with 245+ classes and sliding window support - Add MedSAM-3D: 3D MedSAM with self-sorting memory bank - Add nnU-Net: Self-configuring U-Net for any biomedical dataset - Add inference functions for all new models - Add Gradio UI tabs for each new model - Add checkpoint download logic for new models - Update feature flags for 7 models total - Update README with comprehensive documentation - Add nnunetv2, medim, monai dependencies6481148 -
Fix short_description length
59b1fc4 -
Add TractSeg support, feature flags, and comprehensive NIfTI support
- Add TractSeg model for white matter bundle segmentation (72 bundles) - Add feature flags via environment variables to enable/disable models: - ENABLE_MEDSAM2, ENABLE_MCP_MEDSAM, ENABLE_TRACTSEG, ENABLE_NEUROSAM3 - Add NeuroSAM3 placeholder (pending model URL from user) - Add comprehensive file format support: - NIfTI (.nii.gz) with affine/header handling - NumPy (.npy) - PIL formats (.png, .jpg) - DICOM (.dcm) - Add nibabel dependency for NIfTI support - Add load_nifti(), save_nifti(), load_image() utilities - Update API to route to appropriate model based on format - Add TractSeg bundle names (72 bundles) - Update README with comprehensive documentation - Rename project to NeuroSeg Servere594362 -
Fix short_description length
9d4d118 -
Add MCP-MedSAM support alongside MedSAM2
- Add MCP-MedSAM 2D segmentation model with modality/content prompts - Keep MedSAM2 for 3D volume segmentation with bi-directional propagation - Add new API endpoints: - segment_mcp_medsam: 2D image segmentation - segment_dicom_mcp_medsam: DICOM 2D segmentation - segment_medsam2: 3D volume segmentation (renamed) - segment_dicom_medsam2: DICOM 3D segmentation (renamed) - Add SimpleMCPMedSAM model implementation - Update Gradio UI with separate tabs for both models - Add transformers dependency for CLIP support - Update README with comprehensive documentation - Support 11 imaging modalities: CT, MRI, PET, X-ray, Ultrasound, etc.7b808c1 -
Refactor: Add MCP server support and improve API documentation
- Add gradio[mcp] dependency for MCP server functionality - Enable mcp_server=True in demo.launch() - Add proper docstrings to all API functions for MCP tool conversion - Add missing config file (sam2.1_hiera_t512.yaml) - Fix PIL deprecation warning (BILINEAR -> Resampling.BILINEAR) - Fix path handling for config and checkpoint files - Add scipy, hydra-core, omegaconf, iopath dependencies - Update README with MCP usage instructionsa1c38b4
-
Add mcp_server=True to demo.launch()
4faae07 -
MedSAM2 Gradio server with ZeroGPU
Matheus Rech- Gradio app with @spaces.GPU(duration=120) for free A10G - Endpoints: segment, segment_dicom, parse_dicom, health - Downloads MedSAM2 checkpoint from wanglab/MedSAM2 on HF Hub - Bidirectional 3D propagation with SAM2.1-tiny backbone - Returns gzip-compressed base64 binary masksece1026 -
initial commit
41833af