File size: 10,101 Bytes
ffbfad7 18e9b8f ffbfad7 18e9b8f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 | """CrossKEY HuggingFace Space -- Interactive 3D Keypoint Matching Demo.
Two-tab Gradio app:
Tab 1 (Explore): Pre-computed results with adjustable matching parameters.
Tab 2 (Your Data): Upload volumes + checkpoint for live inference.
"""
import logging
import os
import sys
# Add space/ to path so local imports work both locally and on HF
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
import gradio as gr
import numpy as np
from inference import load_precomputed, run_inference, run_matching
from visualization import build_matching_figure
# ZeroGPU decorator -- no-op when running locally
try:
import spaces
gpu_decorator = spaces.GPU
except ImportError:
gpu_decorator = lambda fn: fn
logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
logger = logging.getLogger("crosskey.app")
# -- Load pre-computed data on startup --
logger.info("Loading pre-computed demo data...")
DEMO_DATA = load_precomputed(
os.path.join(os.path.dirname(os.path.abspath(__file__)), "precomputed")
)
logger.info(
"Loaded: %d MR descriptors, %d US descriptors",
len(DEMO_DATA["descriptors_mr"]),
len(DEMO_DATA["descriptors_us"]),
)
def update_demo(
ratio_threshold: float,
evaluation_threshold: float,
mutual: bool,
metric: str,
) -> tuple:
"""Re-run matching with new parameters and rebuild the figure."""
match_pairs, metrics = run_matching(
DEMO_DATA["descriptors_mr"],
DEMO_DATA["descriptors_us"],
DEMO_DATA["points_mr"],
DEMO_DATA["points_us"],
ratio_threshold=ratio_threshold,
mutual=mutual,
metric=metric,
evaluation_threshold=evaluation_threshold,
)
fig = build_matching_figure(
volume_mr=DEMO_DATA["volume_mr"],
volume_us=DEMO_DATA["volume_us"],
points_mr=DEMO_DATA["points_mr"],
points_us=DEMO_DATA["points_us"],
padded_shape_mr=tuple(DEMO_DATA["metadata"]["padded_shape_mr"]),
padded_shape_us=tuple(DEMO_DATA["metadata"]["padded_shape_us"]),
match_pairs=match_pairs,
metrics=metrics,
evaluation_threshold=evaluation_threshold,
)
return (
fig,
metrics['num_matches'],
metrics['num_correct'],
round(metrics['precision'], 1),
round(metrics['matching_score'], 4),
)
@gpu_decorator
def run_custom_inference(mr_file, us_file, heatmap_file, ckpt_file):
"""Run inference on uploaded data. Uses ZeroGPU on HF Spaces."""
if any(f is None for f in [mr_file, us_file, heatmap_file, ckpt_file]):
raise gr.Error("Please upload all four files: MR volume, US volume, heatmap, and checkpoint.")
logger.info("Running inference on uploaded data...")
data = run_inference(
mr_path=mr_file.name,
us_path=us_file.name,
heatmap_path=heatmap_file.name,
checkpoint_path=ckpt_file.name,
)
return data
def update_custom(
data: dict,
ratio_threshold: float,
evaluation_threshold: float,
mutual: bool,
metric: str,
) -> tuple:
"""Re-run matching on custom data with new parameters."""
if data is None:
raise gr.Error("Run inference first.")
match_pairs, metrics = run_matching(
data["descriptors_mr"],
data["descriptors_us"],
data["points_mr"],
data["points_us"],
ratio_threshold=ratio_threshold,
mutual=mutual,
metric=metric,
evaluation_threshold=evaluation_threshold,
)
fig = build_matching_figure(
volume_mr=data["volume_mr"],
volume_us=data["volume_us"],
points_mr=data["points_mr"],
points_us=data["points_us"],
padded_shape_mr=tuple(data["metadata"]["padded_shape_mr"]),
padded_shape_us=tuple(data["metadata"]["padded_shape_us"]),
match_pairs=match_pairs,
metrics=metrics,
evaluation_threshold=evaluation_threshold,
)
return (
fig,
metrics['num_matches'],
metrics['num_correct'],
round(metrics['precision'], 1),
round(metrics['matching_score'], 4),
)
# -- Build Gradio UI --
with gr.Blocks(
title="CrossKEY -- 3D Cross-modal Keypoint Matching",
theme=gr.themes.Soft(),
css="footer {display: none !important;}",
) as demo:
gr.Markdown(
"# CrossKEY\n"
"**3D Cross-modal Keypoint Descriptor for MR-US Matching and Registration**"
)
with gr.Tabs():
# ---- Tab 1: Explore ----
with gr.Tab("Explore"):
with gr.Row():
with gr.Column(scale=1, min_width=260):
gr.Markdown("### Matching Parameters")
demo_ratio = gr.Slider(
minimum=0.5, maximum=1.0, value=0.75, step=0.05,
label="Ratio Threshold",
)
demo_eval_thresh = gr.Slider(
minimum=1.0, maximum=10.0, value=5.0, step=0.5,
label="Evaluation Threshold (mm)",
)
demo_mutual = gr.Checkbox(value=True, label="Mutual Nearest Neighbor")
demo_metric = gr.Dropdown(
choices=["euclidean", "cosine"], value="euclidean",
label="Distance Metric",
)
gr.Markdown("### Results")
with gr.Group():
with gr.Row():
demo_n_matches = gr.Number(label="Matches", interactive=False)
demo_n_correct = gr.Number(label="Correct", interactive=False)
with gr.Row():
demo_precision = gr.Number(label="Precision (%)", interactive=False)
demo_match_score = gr.Number(label="Match Score", interactive=False)
with gr.Column(scale=3):
demo_plot = gr.Plot(label="3D Matching Visualization")
demo_inputs = [demo_ratio, demo_eval_thresh, demo_mutual, demo_metric]
demo_outputs = [demo_plot, demo_n_matches, demo_n_correct, demo_precision, demo_match_score]
# Update on any parameter change
for inp in demo_inputs:
inp.change(fn=update_demo, inputs=demo_inputs, outputs=demo_outputs)
# Load initial results
demo.load(fn=update_demo, inputs=demo_inputs, outputs=demo_outputs)
# ---- Tab 2: Your Data ----
with gr.Tab("Your Data"):
gr.Markdown(
"Upload your own MR volume, US volume, heatmap, and a trained CrossKEY checkpoint.\n\n"
"Inference runs on GPU and may take 30-60 seconds."
)
with gr.Row():
custom_mr = gr.File(label="MR Volume (.nii.gz)", file_types=[".nii.gz"])
custom_us = gr.File(label="US Volume (.nii.gz)", file_types=[".nii.gz"])
with gr.Row():
custom_heatmap = gr.File(label="Heatmap (.nii.gz)", file_types=[".nii.gz"])
custom_ckpt = gr.File(label="Checkpoint (.ckpt)", file_types=[".ckpt"])
custom_run_btn = gr.Button("Run Inference (GPU)", variant="primary")
with gr.Row():
with gr.Column(scale=1, min_width=260):
gr.Markdown("### Matching Parameters")
custom_ratio = gr.Slider(
minimum=0.5, maximum=1.0, value=0.75, step=0.05,
label="Ratio Threshold",
)
custom_eval_thresh = gr.Slider(
minimum=1.0, maximum=10.0, value=5.0, step=0.5,
label="Evaluation Threshold (mm)",
)
custom_mutual = gr.Checkbox(value=True, label="Mutual Nearest Neighbor")
custom_metric = gr.Dropdown(
choices=["euclidean", "cosine"], value="euclidean",
label="Distance Metric",
)
gr.Markdown("### Results")
with gr.Group():
with gr.Row():
custom_n_matches = gr.Number(label="Matches", interactive=False)
custom_n_correct = gr.Number(label="Correct", interactive=False)
with gr.Row():
custom_precision = gr.Number(label="Precision (%)", interactive=False)
custom_match_score = gr.Number(label="Match Score", interactive=False)
with gr.Column(scale=3):
custom_plot = gr.Plot(label="3D Matching Visualization")
# State to hold inference results
custom_data_state = gr.State(value=None)
custom_param_inputs = [custom_ratio, custom_eval_thresh, custom_mutual, custom_metric]
custom_outputs = [custom_plot, custom_n_matches, custom_n_correct, custom_precision, custom_match_score]
# Inference button: run model, then update visualization
def infer_and_display(mr_file, us_file, heatmap_file, ckpt_file, ratio, eval_thresh, mutual, metric):
data = run_custom_inference(mr_file, us_file, heatmap_file, ckpt_file)
fig, n_m, n_c, prec, ms = update_custom(data, ratio, eval_thresh, mutual, metric)
return data, fig, n_m, n_c, prec, ms
custom_run_btn.click(
fn=infer_and_display,
inputs=[custom_mr, custom_us, custom_heatmap, custom_ckpt] + custom_param_inputs,
outputs=[custom_data_state] + custom_outputs,
)
# Re-match on parameter change (no re-inference)
for inp in custom_param_inputs:
inp.change(
fn=update_custom,
inputs=[custom_data_state] + custom_param_inputs,
outputs=custom_outputs,
)
if __name__ == "__main__":
demo.launch()
|