code for reproducing Fig. 7 in the paper

Of course. The figure 7 is the polar map comparison of images before and after alignment.

If you want the alignment code that can achieve the performance, it's the 'homography_alignment.py' provided in our repository.
If you want the polar map visualization code, we first use the pretrained pwc-network to estimate the flow among each frame and the base frame. Assuming that you have got the 13 flow maps for each burst, then we visualize these flows with polar map.

import numpy as np
import matplotlib.pyplot as plt
import flowpy
import os
import cmath

save_path = 'Manual_polar_map_flo_fix20'
if not os.path.exists(save_path):
    os.makedirs(save_path)

flo_path = '../pytorch-pwc-master/result/patch/LR_patch_arrow/baseframe/test_LR_flo'
test_LR_path = '../pytorch-pwc-master/result/patch/LR_patch_arrow/baseframe/test_LR'

colormap = ['#FF0000', '#00FF00', '#0000FF', '#00FFFF', '#FF00FF', '#FFFF00', '#000000', '#A2142F',
            '#0072BD', '#D95319', '#EDB120', '#7E2F8E', '#77AC30']

def draw_polar(LR_list):
    fig = plt.figure()
    for number in range(1, 14):
        flo_name = 'out_{}_{}.flo'.format(LR_list, number)
        flo = flowpy.flow_read(os.path.join(flo_path, flo_name))
        height, width, co = flo.shape

        x_plane = flo[:, :, 0].reshape(-1)
        y_plane = flo[:, :, 1].reshape(-1)

        length = len(x_plane)
        r_polar = []
        theta_polar = []
        for i in range(length):
            cor = complex(x_plane[i], y_plane[i])
            r, theta = cmath.polar(cor)
            r_polar.append(r)
            theta_polar.append(theta)

        ax = fig.add_subplot(111, projection='polar')

        ax.set_ylim(0, 20)
        ax.set_yticks(np.arange(0, 20, 1.0))

        c = ax.scatter(theta_polar, r_polar, c=colormap[number-1], s=1, cmap='hsv', alpha=0.75)

    save_name = '{}_polar_map.png'.format(LR_list)
    plt.savefig(os.path.join(save_path,save_name))
    plt.close('all')

yjsunnn / FBANet

code for reproducing Fig. 7 in the paper #4