#!/usr/bin/python
from common import *
from pathlib import Path

fn = sys.argv[1]
a = slice(fn)
fn = Path(fn).stem
for i in range(len(a)):
  cv.imwrite(f'data/{fn}-{i}.png', a[i])

exit()

from PIL import Image, ImageDraw
img = Image.open(sys.argv[1])
id = ImageDraw.Draw(img)
sz = img.size; S,L,T = size(sz[0], sz[1])
for yy in range(10):
  for xx in range(9):
    y = T + yy * S
    x = L + xx * S
    id.rectangle((x,y,x+S,y+S), outline=(0,255,0,255), width=1)
img.show()

#!/usr/bin/python
from glob import glob
from shutil import move

s = 'rnbakcp'; s += s.upper() + '0'

for c in s:
 for p in [c + '*.png', c + '*.jpg']:
  for f in glob(p): move(f, c)

from common import *
from sklearn.svm import SVC
import pickle

data_dir = "data"
features = []; labels = []
for label in os.listdir(data_dir):
  label_dir = os.path.join(data_dir, label)
  if not os.path.isdir(label_dir): continue
  for img_file in os.listdir(label_dir):
    img_path = os.path.join(label_dir, img_file)
    img = cv.imread(img_path)
    if img is None: continue
    feature = extract_features(img)
    if feature is None: continue
    features.append(feature)
    labels.append(label)

model = SVC(kernel='rbf', C=10, gamma=0.1)
model.fit(features, labels)
with open("model.pkl", "wb") as f: pickle.dump(model, f)

import numpy as np
import cv2 as cv
import pickle
import os
import sys

def size (W, H):
  S = 68 if W == 640 else 86 # size
  L = (W - 9 * S) // 2 # left
  T = (H - 10 * S) // 2 # top
  return S,L,T

def slice (fn):
  all = []

  img = cv.imread(fn); H,W = img.shape[:2]
  S,L,T = size(W, H)

  mask = np.zeros((S, S), dtype=bool)
  HS = S // 2; R = HS - 4 # half S & radius
  for y in range(S):
    for x in range(S):
      # 圓形mask. hypotenuse: 斜邊
      if np.hypot(x - HS, y - HS) > R: mask[y, x] = True

  for yy in range(10):
    for xx in range(9):
      x = L + xx * S; y = T + yy * S
      x2 = min(W, x + S); y2 = min(H, y + S)
      sl = img[y:y2, x:x2]; sl[mask] = 0
      all.append(sl)

  return all

def extract_features (img):
  gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
  _, bin = cv.threshold(gray, 0, 255, cv.THRESH_BINARY + cv.THRESH_OTSU)
  contours, _ = cv.findContours(bin, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
  if len(contours) == 0: return None

  contour = max(contours, key=cv.contourArea)
  moments = cv.moments(contour)
  hu_moments = cv.HuMoments(moments).flatten()

  # 在陰影或強光條件下，RGB會整體變化，而HSV的色調H和飽和度S相對穩定，更適合顏色識別任務。V是亮度。
  hsv = cv.cvtColor(img, cv.COLOR_BGR2HSV)
  color_features = np.concatenate([
    cv.mean(hsv)[:3],
    cv.meanStdDev(hsv)[1].flatten()
  ])

  return np.concatenate([hu_moments, color_features])

#!/usr/bin/python
import numpy as np
import cv2 as cv
import sys
import os

fn = sys.argv[1]
img = cv.imread(fn); height, width = img.shape[:2]
fn = os.path.basename(fn)

circles = cv.HoughCircles(
  cv.cvtColor(img, cv.COLOR_BGR2GRAY), # 只支持灰度
  cv.HOUGH_GRADIENT, # 梯度，not 漸變
  1, # 改成2影響很大
  60, # Minimum distance between the centers of the detected circles
  param1=150, # 減小可增加檢測到的圓數量
  param2=35, # The smaller it is, the more false circles may be detected.
  minRadius=30,
  maxRadius=48
) # 返回shape為(1,n,3)的ndarray
if circles is None: exit()
circles = np.around(circles).astype(int)

mask = np.zeros((96,96), dtype=bool)
for y in range(96):
  for x in range(96):
    # 以(48,48)為圓心，40為半徑的mask. hypotenuse 斜邊
    if np.hypot(x - 48, y - 48) > 40: mask[y, x] = True

for x,y,r in circles[0]:
  cv.circle(img, (x, y), r, (0, 255, 0), 2)

  x1 = max(0, x - r); x2 = min(width, x + r)
  y1 = max(0, y - r); y2 = min(height, y + r)
  ps = cv.resize(img[y1:y2, x1:x2], (96,96))
  ps[mask] = 255
  #cv.imwrite(f'tmp/{fn}-{x},{y}.png', ps)

from PIL import Image
img = Image.fromarray(cv.cvtColor(img, cv.COLOR_BGR2RGB))
img.show()

print(f'{circles.shape[1]} pieces')