Реши. Выбрать нужно вариант чисел от 1 до 15

python
from PIL import Image, ImageOps, ImageFilter, ImageEnhance, ImageStat, ImageDraw, ImageFont, Image
img2=Image.open('/mnt/data/IMG_20251024_133046.jpg')
img2.size


(4624, 3472)

python
img2_cropped=img2.crop((1500,1100,4400,3200)).resize((900,800))
img2_cropped


<PIL.Image.Image image mode=RGB size=900x800>

python
display(img2_cropped.crop((410,380,880,720)).resize((600,400)))

python
display(img2_cropped.crop((0,200,450,650)).resize((600,600)))

python
display(img2_cropped.crop((150,0,500,250)).resize((600,400)))

python
display(img2_cropped.crop((0,40,300,280)).resize((600,400)))

python
display(img2_cropped.crop((530,40,880,400)).resize((600,400)))

python
display(img2_cropped.crop((350,450,640,720)).resize((500,400)))

python
display(img2_cropped.crop((330,330,700,580)).resize((500,400)))

python
display(img2_cropped.crop((410,180,540,350)).resize((250,300)))

python
display(img2_cropped.crop((300,260,800,650)).resize((700,500)))

python
import math, itertools, collections, numpy as np, pandas as pd

# Define training points as (x,y,class)
train = []
# Orange class 3 coords:
for p in [(1,2),(2,1),(2,0),(4,2),(4,1),(6,1)]:
    train.append((*p,3))
# Blue class 1 (except test (3,4))
for p in [(2,3),(4,3),(5,3)]:
    train.append((*p,1))
# Red class 0 except test (8,2)
for p in [(8,3),(9,2)]:
    train.append((*p,0))
# Green class 2 except test (5,0)
for p in [(6,0),(7,0),(7,-1),(8,-1)]:
    train.append((*p,2))
len(train), train[:5]


(15, [(1, 2, 3), (2, 1, 3), (2, 0, 3), (4, 2, 3), (4, 1, 3)])

python
tests = [((8,2),0), ((3,4),1), ((5,0),2)]

def knn_predict(k, point):
    x,y = point
    # compute squared Euclidean distances
    dist = [((x-xi)**2 + (y-yi)**2, cls) for xi,yi,cls in train]
    # sort by distance then class number (tie rule for nearest ordering)
    dist.sort(key=lambda t: (t[0], t[1]))
    # get first k neighbors
    k_neighbors = dist[:k]
    # count frequencies
    freq = collections.Counter(cls for d,cls in k_neighbors)
    # find max frequency
    max_freq = max(freq.values())
    candidate_classes = [cls for cls,count in freq.items() if count==max_freq]
    if len(candidate_classes)==1:
        return candidate_classes[0]
    # tie: compute sum of distances among neighbors for each class
    sums = {cls: sum(d for d, c in k_neighbors if c==cls) for cls in candidate_classes}
    min_sum = min(sums.values())
    candidates = [cls for cls, s in sums.items() if s==min_sum]
    if len(candidates)==1:
        return candidates[0]
    # still tie -> choose smaller class number
    return min(candidates)

# For k=1..15 predictions for each test:
results = {}
for k in range(1,16):
    preds=[(point,knn_predict(k,point)) for point,_ in tests]
    correct = [pred==true for (point,true),(_,pred) in zip(tests, preds)]
    results[k]= (preds, correct)
results[1]


([((8, 2), 0), ((3, 4), 1), ((5, 0), 2)], [True, True, True])

python
good_k=[]
details={}
for k,(preds,correct) in results.items():
    if all(correct):
        good_k.append(k)
    details[k]=preds,correct
good_k, details[2]


([1, 2, 5], ([((8, 2), 0), ((3, 4), 1), ((5, 0), 2)], [True, True, True]))

python
[(k, all(res[1])) for k,res in details.items()]


[(1, True),
 (2, True),
 (3, False),
 (4, False),
 (5, True),
 (6, False),
 (7, False),
 (8, False),
 (9, False),
 (10, False),
 (11, False),
 (12, False),
 (13, False),
 (14, False),
 (15, False)]

python
for k in range(1,16):
    preds,_=details[k]
    print(k, [(point,pred) for point,pred in preds], 'allcorrect', all(_))

python
import math
def knn_predict(k, point):
    x,y = point
    dist = [ (math.dist((x,y),(xi,yi)), cls) for xi,yi,cls in train]
    dist.sort(key=lambda t: (t[0], t[1]))
    k_neighbors = dist[:k]
    # counts
    freq = collections.Counter(cls for d,cls in k_neighbors)
    max_freq = max(freq.values())
    candidate_classes = [cls for cls,count in freq.items() if count==max_freq]
    if len(candidate_classes)==1:
        return candidate_classes[0]
    sums = {cls: sum(d for d,c in k_neighbors if c==cls) for cls in candidate_classes}
    min_sum = min(sums.values())
    candidates = [cls for cls,s in sums.items() if s == min_sum

[(1, True, [((8, 2), 0), ((3, 4), 1), ((5, 0), 2)]),
 (2, True, [((8, 2), 0), ((3, 4), 1), ((5, 0), 2)]),
 (3, False, [((8, 2), 0), ((3, 4), 1), ((5, 0), 3)]),
 (4, False, [((8, 2), 0), ((3, 4), 1), ((5, 0), 3)]),
 (5, True, [((8, 2), 0), ((3, 4), 1), ((5, 0), 2)]),
 (6, False, [((8, 2), 2), ((3, 4), 1), ((5, 0), 3)]),
 (7, False, [((8, 2), 2), ((3, 4), 3), ((5, 0), 3)]),
 (8, False, [((8, 2), 2), ((3, 4), 3), ((5, 0), 3)]),
 (9, False, [((8, 2), 2), ((3, 4), 3), ((5, 0), 3)]),
 (10, False, [((8, 2), 2), ((3, 4), 3), ((5, 0), 3)]),
 (11, False, [((8, 2), 2), ((3, 4), 3), ((5, 0), 3)]),
 (12, False, [((8, 2), 2), ((3, 4), 3), ((5, 0), 3)]),
 (13, False, [((8, 2), 2), ((3, 4), 3), ((5, 0), 3)]),
 (14, False, [((8, 2), 3), ((3, 4), 3), ((5, 0), 3)]),
 (15, False, [((8, 2), 3), ((3, 4), 3), ((5, 0), 3)])]