diff --git a/img2nc.py b/img2nc.py
new file mode 100644
index 0000000000000000000000000000000000000000..dd47db8a93bce590363ee87e49db73e3de945f4a
--- /dev/null
+++ b/img2nc.py
@@ -0,0 +1,225 @@
+import math
+import numpy as np
+import matplotlib.pyplot as plt
+import cv2
+
+
+def imshow(img, w_max=1200, h_max=700, name="show"):
+ h = img.shape[0]
+ w = img.shape[1]
+ ratio = w / h
+ ratio_max = w_max / h_max
+
+ if ratio >= ratio_max:
+ w_new = w_max
+ h_new = int(h * w_new / w)
+ else:
+ h_new = h_max
+ w_new = int(w * h_new / h)
+ img_show = cv2.resize(img, (w_new, h_new))
+ cv2.imshow(name, img_show)
+
+
+def mill(img_mill, img_keep, dpi, dia_mm, stepover=0.8, once=False):
+ global count
+ img_mill_bin = np.zeros_like(img_mill)
+ img_mill_bin[img_mill >= 127] = 255
+
+ img_keep_bin = np.zeros_like(img_keep)
+ img_keep_bin[img_keep >= 127] = 255
+
+ dpmm = dpi / 25.4
+ dia_px = int(dia_mm * dpmm)
+
+ print(f"Tool: {dia_mm} mm = {dia_px} px")
+ if dia_px < 10:
+ print(f"Warning: tool is only {dia_px} pixels wide")
+
+ k_step = int(dia_mm * 2 * stepover * dpmm)
+ k_tool = int(dia_mm * dpmm)
+
+ kernel_step = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (k_step, k_step))
+ kernel_tool = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (k_tool, k_tool))
+
+ done = False
+
+ paths = []
+
+ # mask = cv2.dilate(img_mill_bin, kernel_tool)
+ # mask = cv2.dilate(mask, kernel_tool)
+ img_keep_pad = np.pad(img_keep_bin, pad_width=(k_tool, k_tool), mode="constant", constant_values=0)
+
+ mask = np.copy(img_mill_bin)
+ # cv2.imwrite(f"mask_init_{count:02d}_pre.png", mask)
+ mask = cv2.dilate(mask, kernel_tool)
+ # mask = cv2.dilate(mask, kernel_tool)
+ mask[img_keep_bin > 0] = 0
+ # cv2.imwrite(f"mask_init_{count:02d}.png", mask)
+ # mask[img_keep > 0] = 0
+ mask = np.pad(mask, pad_width=(k_tool, k_tool), mode="constant", constant_values=0)
+
+ h_p, w_p = mask.shape
+
+ edge_px = k_tool
+
+ mask_diff = np.zeros_like(mask)
+ mask_miss = np.zeros_like(mask)
+ mask_next = np.zeros_like(mask)
+
+ i = 0
+ img_miss = np.zeros_like(img_mill_bin)
+
+ while not done:
+ if i == 0:
+ dil = cv2.dilate(img_keep_pad, kernel_tool)
+ mask_next[(mask > 0) & ~(dil > 0)] = 255
+ mask_next_dil = cv2.dilate(mask_next, kernel_tool)
+ missed = (mask > 0) & ~(mask_next_dil > 0)
+ img_miss[missed[edge_px:-edge_px, edge_px:-edge_px]] = 255
+ # ignore pixels that weren't marked to begin with
+ img_miss[img_mill_bin == 0] = 0
+ else:
+ mask_next[:, :] = cv2.erode(mask, kernel_step)
+ mask_diff[~(mask > 0)] = 0
+ mask_diff[mask > 0] = 255
+ mask_diff[mask_next > 0] = 0
+ mask_miss[:, :] = cv2.erode(mask_diff, kernel_tool)
+ contours, hierarchy = cv2.findContours(mask, cv2.RETR_TREE, cv2.CHAIN_APPROX_TC89_KCOS)
+ for c in contours:
+ c_float = c.astype(np.float32)
+ c_mm = np.zeros_like(c_float)
+ c_mm[:, 0, 0] = (c_float[:, 0, 0] - edge_px) / dpmm
+ c_mm[:, 0, 1] = ((h_p - c_float[:, 0, 1]) - edge_px) / dpmm
+ paths.append(c_mm[:, 0, :])
+ if once:
+ done = True
+
+ mask[:, :] = mask_next[:, :]
+ if i > 0:
+ mask[mask_miss > 0] = 255
+
+ i += 1
+
+ if not np.any(mask > 0):
+ done = True
+
+ print(i)
+
+ return paths, img_miss
+
+
+def add_gcode(f, paths, z_down=0.0, z_up=1.0):
+ feedrate_cut = 480.0
+ feedrate_plunge = 240.0
+
+ for p in paths:
+ x = p[0, 0]
+ y = p[0, 1]
+ f.write(f"G0 X{x:.4f} Y{y:.4f}\n")
+ f.write(f"G1 Z{z_down:.4f} F{feedrate_plunge:.4f}\n")
+
+ for c in p:
+ x, y = c[:]
+ f.write(f"G1 X{x:.4f} Y{y:.4f} F{feedrate_cut:.4f}\n")
+
+ x = p[0, 0]
+ y = p[0, 1]
+ f.write(f"G1 X{x:.4f} Y{y:.4f} F{feedrate_cut:.4f}\n")
+ f.write(f"G0 Z{z_up:.4f}\n")
+
+
+def main():
+ filename_copper = "input/Cu.png"
+ filename_edge = "input/Edge.png"
+ filename_out = "result.nc"
+
+ dpi = 1000.0
+
+ img_copper = cv2.imread(filename_copper)
+ img_edge = cv2.imread(filename_edge)
+
+ if len(img_copper.shape) > 2:
+ img_copper = cv2.cvtColor(img_copper, cv2.COLOR_BGR2GRAY)
+
+ if len(img_edge.shape) > 2:
+ img_edge = cv2.cvtColor(img_edge, cv2.COLOR_BGR2GRAY)
+
+ img1 = np.zeros_like(img_edge)
+ img1[(img_edge < 127) | (img_copper > 127)] = 255
+ img1_keep = np.copy(img1)
+ img1_keep[img1_keep >= 127] = 255
+ img1_keep[img1_keep < 127] = 0
+ img1_mill = 255 - img1_keep
+
+ paths1, img_miss1 = mill(img1_mill, img1_keep, 1000, 0.794)
+
+ paths2, img_miss2 = mill(img_miss1, img1_keep, 1000, 0.397)
+
+ img_edge_keep = np.copy(img_edge)
+ img_edge_keep[img_edge_keep >= 127] = 255
+ img_edge_keep[img_edge_keep < 127] = 0
+ img_edge_mill = 255 - img_edge_keep
+ paths3, _ = mill(img_edge_mill, img_edge_keep, dpi, 0.794, once=True)
+
+ f = open(filename_out, "w")
+
+ f.write("%\n")
+ f.write("G17\n")
+ f.write("G21\n")
+ f.write("G40\n")
+ f.write("G49\n")
+ f.write("G54\n")
+ f.write("G80\n")
+ f.write("G90\n")
+ f.write("G94\n")
+ f.write("T4 M06\n")
+ f.write("S16000\n")
+ f.write("G0 Z1\n")
+ f.write("M03\n")
+
+ add_gcode(f, paths2, z_down=-0.12)
+
+ f.write("M05\n")
+ f.write("T3 M06\n")
+ f.write("S16000\n")
+ f.write("G0 Z1\n")
+ f.write("M03\n")
+
+ add_gcode(f, paths1, z_down=-0.12)
+
+ add_gcode(f, paths3, z_down=-0.6)
+ add_gcode(f, paths3, z_down=-1.2)
+ add_gcode(f, paths3, z_down=-1.75)
+
+ f.write("M30\n")
+ f.write("%\n")
+ f.write("M6 T-1\n")
+ f.write("M496.1\n")
+
+ f.close()
+
+ dpmm = dpi / 25.4
+
+ h, w = np.shape(img1_mill)
+ img_show = np.zeros((h, w, 3), dtype=np.uint8)
+ img_show[img1_mill > 0, :] = 150
+ img_show[img_miss2 > 0, :] = 0
+ img_show[img_miss2 > 0, 0] = 255
+
+ plt.figure()
+ plt.imshow(img_show)
+ for c in paths1:
+ plt.plot(c[:, 0] * dpmm, h-c[:, 1] * dpmm, "b")
+ plt.plot([c[-1, 0] * dpmm, c[0, 0] * dpmm], [h-c[-1, 1] * dpmm, h-c[0, 1] * dpmm], "b")
+ for c in paths2:
+ plt.plot(c[:, 0] * dpmm, h-c[:, 1] * dpmm, "g")
+ plt.plot([c[-1, 0] * dpmm, c[0, 0] * dpmm], [h-c[-1, 1] * dpmm, h-c[0, 1] * dpmm], "g")
+ for c in paths3:
+ plt.plot(c[:, 0] * dpmm, h-c[:, 1] * dpmm, "b")
+ plt.plot([c[-1, 0] * dpmm, c[0, 0] * dpmm], [h-c[-1, 1] * dpmm, h-c[0, 1] * dpmm], "b")
+ plt.axis("equal")
+ plt.show()
+
+
+if __name__ == "__main__":
+ main()