Updated gradio app

gradio/app.py

@@ -1,28 +1,68 @@
 import gradio as gr
 import numpy as np
 import cv2
+from matplotlib import pyplot as plt
 from multifocal_stitching import candidate_stitches, stitch, merge
 
-def stitch_interface(img1, img2):
-                     #use_wins, workers, peak_cutoff_std,
-                     #peaks_dist_threshold, filter_radii, min_overlap,
-                     #early_term_thresh):
-    res = stitch(*[cv2.cvtColor(np.array(im), cv2.COLOR_RGB2GRAY) for im in (img1, img2)])
-    dx, dy = res.coord
-    return merge(img1, img2, dx, dy, resize_factor=8)
+def stitch_interface(img1, img2,
+                     filter_radii,
+                     min_overlap,
+                     peak_cutoff_std,
+                     peaks_dist_threshold,
+                     use_wins):
+    grey_imgs = [cv2.cvtColor(np.array(im), cv2.COLOR_RGB2GRAY) for im in (img1, img2)]
+    WINS = [(0,), (1,), (0,1)]
+    results = sorted(list(candidate_stitches(*grey_imgs,
+                                             early_term_thresh=1.0,
+                                             filter_radii=filter_radii,
+                                             min_overlap=min_overlap,
+                                             peak_cutoff_std=peak_cutoff_std,
+                                             peaks_dist_threshold=peaks_dist_threshold,
+                                             use_wins=WINS[use_wins]
+                                             )),
+                     key=lambda r: r.corr_coeff, reverse=True)
+    best = results[0]
+    dx, dy = best.delta
+    table = [[r.corr_coeff, r.area, r.r, r.use_win, int(r.delta[0]), int(r.delta[1])]
+             for r in results]
+    fig = plt.figure()
+    plt.imshow(abs(best.corr))
+    xs, ys = zip(*set(tuple(r.freq_delta) for r in results))
+    plt.scatter(xs, ys, marker='x', c='r')
+    return merge(img1, img2, dx, dy, resize_factor=8), fig, table
 
+examples = [
+    [f'tests/imgs/{name}_1_small.jpg',
+     f'tests/imgs/{name}_2_small.jpg',
+     ] for name in (
+         'high_freq_features',
+         'low_freq_features',
+         'large_overlap',
+         'small_overlap',
+         'sharp_blur_overlap',
+     )
+]
 demo = gr.Interface(fn=stitch_interface, inputs=[
-    gr.Image(type='pil'), gr.Image(type='pil')
-], outputs=[
-    gr.Gallery()
-], examples=[
-    ["tests/imgs/high_freq_features_1_small.jpg",
-     "tests/imgs/high_freq_features_2_small.jpg", ]
-])
-
-
-
-
+    gr.Image(type='pil'), gr.Image(type='pil'),
+    gr.Dropdown(list(range(10,501,10)), value=[100, 50, 20], multiselect=True, label="filter_radii",
+                info="Low-pass filter radii to try, smaller matches coarser/out-of-focus features"),
+    gr.Slider(minimum=0, maximum=1, step=0.001, value=0.125, label='min_overlap',
+              info='Set lower limit for overlapping region as a fraction of total image area'),
+    gr.Slider(minimum=0, maximum=5, step=0.1, value=1.0, label='peak_cutoff_std',
+              info='Number of standard deviations below max value to use for peak finding'),
+    gr.Slider(minimum=1, maximum=100, step=1, value=25, label='peaks_dist_threshold',
+              info='Distance to consider as part of same cluster when finding peak centroid'),
+    gr.Radio(["No window", "Hanning window", "Both"], label='use_wins',type="index",
+             info='Whether to try using Hanning window'),
 
+], outputs=[
+    gr.Gallery(label='Merged Images'),
+    gr.Plot(label='Frequency domain peaks'),
+    gr.DataFrame(label='Candidate Stitches',
+                 headers=['Corr Value', 'Area', 'r', 'use_win', 'X offset', 'Y offset'],
+                 datatype=['number', 'number', 'number', 'number', 'number', 'bool'],
+                 max_rows=10),
+], examples=examples)
 
-demo.launch()
+if __name__ == '__main__':
+    demo.launch()

src/multifocal_stitching/stitching.py

@@ -68,7 +68,7 @@ def get_peak_centroids(res: np.ndarray,
 
 StitchingResult = namedtuple(
     'StitchingResult',
-    ['corr_coeff', 'corr', 'delta', 'val', 'area', 'r', 'use_win']
+    ['corr_coeff', 'corr', 'freq_delta', 'delta', 'val', 'area', 'r', 'use_win']
 )
 
 def print_stitching_result(r: StitchingResult):
@@ -115,7 +115,7 @@ def candidate_stitches(img1: np.ndarray, img2: np.ndarray,
             for dy, dx in get_peak_centroids(res, peak_cutoff_std, peaks_dist_threshold):
                 for dX, dY in product((dx, -X+dx), (dy, -Y+dy)):
                     coef, area = get_overlap(img1, img2, (dX, dY), min_overlap=min_overlap)
-                    result = StitchingResult(coef, res, (dX, dY), res[dY, dX], area, r, use_win)
+                    result = StitchingResult(coef, res, (dx, dy), (dX, dY), res[dY, dX], area, r, use_win)
                     if verbose:
                         print_stitching_result(result)
                     yield result