Mavzu: Qo’lda yozilgan raqamni aniqlash uchun dataset shakllantirish Bajardi : Ismoilov Temurbek Qabul qildi : Temirov Azizbek tuit 2022-2023 Ishning maqsadi

O’ZBEKISTON RESPUBLIKASI AXBOROT TEXNALOGIYALARI VA KOMMUNIKATSIYALARINI RIVOJLANTIRISH VAZIRLIGI
MUHAMMAD AL-XORAZMIY NOMIDAGI
TOSHKENT AXBOROT TEXNOLOGIYALARI UNIVERSITETI

Kafedra : Kompyuter tizimlari

Fan nomi : Ma'lumotlarning intellektual tahlili

3-AMALIY MASHG’ULOT

Mavzu: Qo’lda yozilgan raqamni aniqlash uchun dataset shakllantirish


Bajardi : Ismoilov Temurbek
Qabul qildi : Temirov Azizbek


TUIT 2022-2023
Ishning maqsadi : Datasetlar bilan ishlar, ularni yaratish va tushunish
Nazariy ma’lumot. Foydalanilgan kodlar tahlili
Bu ishni bajarishimiz uchun KNN algoritmidan foydalanamiz

K ta eng yaqin qo’shni sonlarni ajratib olish va o’zaro taqqoslash asosida sonlarni topamiz

Code		Ta’rifi
from PIL import Image import matplotlib.pyplot as plt import numpy as np		Biz avvalo kerakli kutubxonalarni yuklaymiz va import qilib olamiz
input_folder = '../input/' filename = input_folder+'digits.png' image = Image.open(filename)		Rasmlarni saqlash uchun papka manzilini tanlaymiz va bizga kerakli rasmlarni ochishni boshlaymiz
def show(img, figsize=(8, 4), title=None): plt.figure(figsize=figsize) plt.imshow(img) if title: plt.title(title) plt.show()		Show() funksiyasini yaratamiz va u bizga kerakli rasmlarni ochib beradi
print(f'Image details') print(f'Format {image.format}') print(f'Size {image.size}') print(f'Mode {image.mode}')		Rasmni tahlil qilamiz
from PIL import ImageEnhance bw_image = image.convert(mode='L') show(bw_image, figsize=(12, 12)) bw_image = ImageEnhance.Contrast(bw_image).enhance(1.5) show(bw_image, figsize=(12, 12))		ImageEnhance funksiyasini import qilib olamiz va rasmni ko’rib chiqamiz
f = plt.figure(figsize=(18,2)) ax = f.subplots(1, 6) for i, n in enumerate(np.random.randint(0, len(samples), size=6)): m = (np.random.randint(0, len(samples[n]))) ax[i].imshow(samples[n][m]) ax[i].set_title(f'Digit: [{n}]') plt.show()		Tanlangan rasmlarni alohida chiqarib ularnitahlil qilamiz
preview = Image.new('L', (len(samples[0])*new_size, len(samples)*new_size)) x = 0 y = 0 for row in resized_samples: for sample in row: preview.paste(sample, (x, y)) x += new_size y+=new_size x = 0 show(preview, figsize=(18,18), title='Processed images') preview.save('preview.png')		Kiritilgan sonlarni chiqarish va saqlashni boshlaymiz
from PIL import ImageOps import matplotlib.patches as patches inv_sample = ImageOps.invert(sample) bbox = inv_sample.getbbox() fig = plt.figure(figsize=(2, 2)) ax = fig.add_axes([0,0,1,1]) ax.imshow(inv_sample) rect = patches.Rectangle( (bbox[0], bbox[3]), bbox[2]-bbox[0], -bbox[3]+bbox[1]-1, fill=False, alpha=1, edgecolor='w') ax.add_patch(rect) plt.show()		Tanlangan raqamni alohida belgilaymiz va aniqlab chiqaramiz
binary_samples = np.array([[sample.getdata() for sample in row] for row in resized_samples]) binary_samples = binary_samples.reshape(len(resized_samples)*len(resized_samples[0]), 28, 28) xfile = 'digits_x_test.npy' yfile = 'digits_y_test.npy' np.save(xfile, binary_samples) np.save(yfile, classes) x_test = np.load(xfile) y_test = np.load(yfile) x_test.shape, y_test.shape for i in np.random.randint(x_test.shape[0], size=6): show(x_test[i], title=f'Digit [{y_test[i]}]', figsize=(1,1))		Hosil qilingan datasetdan malumotlarni olamiz va tekshiramiz