added annotation script

extra/annotation_ui.py

@@ -0,0 +1,164 @@
+"""
+Starts a napari widget for annotating points on an image starting from LoG candidates or existing results stored in a CSV.
+
+The widget allows to:
+- toggle points visibility
+- save points to a csv file
+- detect points using LoG detector
+- increase/decrease LoG threshold with w/e keys and a slider
+
+Usage:
+    python annotation_ui.py <path_to_image> [options]
+"""
+import numpy as np
+import napari
+from pathlib import Path
+from tifffile import imread
+from qtpy.QtWidgets import QMessageBox
+from csbdeep.utils import normalize
+from skimage.feature import blob_log
+from magicgui import magicgui
+import pandas as pd
+import argparse
+from spotiflow.utils import read_coords_csv
+from spotiflow.utils.fitting import estimate_params
+
+
+KEY_SHORTCUTS = {
+    'toggle':  ('q', 'toggle points'),
+    'save':    ('s' , 'save csv'),
+    'thr_dec': ('w', 'decrease thr'),
+    'thr_inc': ('e', 'increase thr'),
+    'detect':  ('d', 'detect')
+    }
+
+def load_points(path):
+    path = Path(path)
+    if path.suffix==".npy":
+        return np.load(args.points)
+    elif path.suffix==".csv":
+        return read_coords_csv(path)
+    else:
+        raise ValueError(f'not supported extension {path.suffix}')
+
+
+def save_points(path, arr):
+    path = Path(path)
+    print(path)
+    if path.suffix==".npy":
+        np.save(path, arr)
+    elif path.suffix==".csv":
+        pd.DataFrame(arr, columns=['y','x']).to_csv(path, index=False)
+    else:
+        raise ValueError(f'not supported extension {path.suffix}')
+
+def filter_points_bbox(points, bbox):
+    """   bbox = ((y1,x1), (y2, x2))  """
+    inds = np.bitwise_and(
+        np.bitwise_and(points[:,0]>=bbox[0,0],points[:,0]<bbox[1,0]),
+        np.bitwise_and(points[:,1]>=bbox[0,1],points[:,1]<bbox[1,1]))
+    return inds
+
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(description='')
+    parser.add_argument("img", type=str)
+    parser.add_argument('-p', "--points", type=str, default=None)
+    parser.add_argument("-t", "--threshold", type=float, default=0.5)
+
+    args = parser.parse_args()
+
+    if args.points is None:
+        args.points = Path(args.img).with_suffix('.csv')            
+
+    img = normalize(imread(args.img), .1,99)
+
+    try:
+        p = load_points(args.points)
+    except Exception as e:
+        print(e)
+        print('using log detector')
+        p = blob_log(img, min_sigma=1, max_sigma=4, threshold=args.threshold,  exclude_border=1)[:,:2]
+        print(p.shape)
+
+    viewer = napari.Viewer()                        
+
+    image_layer = viewer.add_image(img, name=Path(args.img).name)
+    image_layer.contrast_limits = [0,3]
+    points_layer = viewer.add_points(p, name='points', border_color='springgreen', border_width=.1, face_color=[0,0,0,0], symbol="o", opacity=.8, size=12)
+
+    # overwrite existing annotation
+    @viewer.bind_key(KEY_SHORTCUTS["save"][0])
+    def f(event=None):
+        res = QMessageBox().warning(
+                    viewer.window.qt_viewer,
+                    "Confirm overwrite",
+                    f"Overwrite existing annotation {args.points.name} ?",
+                    QMessageBox.Yes | QMessageBox.No,
+                    QMessageBox.No)
+        if res == QMessageBox.Yes:
+            save_points(args.points, points_layer.data)
+            print(points_layer.data)
+            print('saved')
+
+    @viewer.bind_key(KEY_SHORTCUTS["toggle"][0], overwrite=True)
+    def f(event=None):
+        if points_layer.visible:
+            points_layer._mymode = points_layer.mode 
+
+        points_layer.visible = not points_layer.visible
+
+        if points_layer.visible:
+            points_layer.mode = points_layer._mymode
+
+    def detect_and_populate():
+        fov = image_layer.corner_pixels
+        ss = slice(fov[0,0], fov[1,0]), slice(fov[0,1], fov[1,1])
+        x = img[ss] 
+        points = blob_log(x, min_sigma=1, max_sigma=parameters.max_sigma.value, threshold=parameters.threshold.value, exclude_border=1)[:,:2]
+        points = points + fov[0]
+        points0 = points_layer.data  
+        idx = filter_points_bbox(points0, fov)
+        points = np.concatenate((points0[~idx], points), axis=0)
+        points_layer.data = points
+        points_layer.selected_data = {}
+
+
+    @magicgui(auto_call=True, max_sigma={"widget_type": "FloatSlider", "min" : 1, "max": 6}, 
+                              threshold={"widget_type": "FloatSlider", "min" : 1e-5, "max": 1, 'step':1000},
+                              fit_gaussian={"widget_type": "CheckBox", "label": "fit gaussian"})
+    def parameters(threshold: float = args.threshold, max_sigma=4, fit_gaussian=False):
+        if not fit_gaussian:
+            detect_and_populate()
+        else:
+            points = points_layer.data
+            x = img
+            params = estimate_params(img=x, centers=points, refine_centers=True, verbose=False)
+            offsets = np.vstack([params.offset_y, params.offset_x]).T
+            print(f"Could not fit {(params.offset_y==np.nan).sum()} points (out of {len(params.offset_y)})")
+            points = points + np.nan_to_num(offsets, 0.0)
+            points_layer.data = points
+
+
+    parameters.threshold._widget._readout_widget.setDecimals(3)
+    viewer.window.add_dock_widget(parameters)
+
+
+    @viewer.bind_key(KEY_SHORTCUTS["detect"][0], overwrite=True)
+    def f(event=None):
+        detect_and_populate()
+
+    @viewer.bind_key(KEY_SHORTCUTS["thr_inc"][0], overwrite=True)
+    def f(event=None):
+        parameters.threshold.value *= 1.05
+
+    @viewer.bind_key(KEY_SHORTCUTS["thr_dec"][0], overwrite=True)
+    def f(event=None):
+        parameters.threshold.value *= 1/1.05
+
+
+
+    viewer.title = " ".join(f"[{v[0]}] {v[1]}   " for k,v in KEY_SHORTCUTS.items())
+
+    napari.run()