Tensorflow datasets: Cats_vs_dogs dataseti

1. Cats vs Dogs dataseti haqida qisqacha ma'lumot

Cats vs Dogs dataseti tensorflow datasetlari ichida binary classification uchun qo'llaniladigan RGB kanalli turli o'lchamdagi 23262 ta tasvirdan iborat tasvirlar to'plami hisoblanadi. Dataset parametrlari:

• Versiya: 4.0.0
• Manba: https://www.tensorflow.org/datasets/catalog/cats_vs_dogs
• Manba kodi: tfds.image_classification.CatsVsDogs
• Dataset hajmi: 786.68 MB

2. Loyihani yuklab olish uchun quyidagi ketma-ketlikni bajaring:

• windows+R klavishlarini bosing va paydo bo'lgan oynaga cmd buyrug'ini yozing OK tugmachasini bosing.

• Loyihani quyidagi link yordamida yuklab oling. (Loyiha uchun yaratilgan fayl adresni o'zingiz ko'rsatishingiz mumkin)

  C:\> git clone https://github.com/MisterFoziljon/CATS_vs_DOGS.git
  

• Loyiha joylashgan faylga kiring.

  C:\> cd CATS_vs_DOGS
  

3. Proyektni ishlatish uchun kerakli modullarni virtual environment yaratib o'rnatib oling.

• O'zingizdagi pip ni so'nggi versiyasiga yangilang.

    C:\CATS_vs_DOGS> python -m pip install --upgrade pip
  

• virtual environment yaratish uchun virtualenv modulini o'rnating.

    C:\CATS_vs_DOGS> python -m pip install --user virtualenv
  

• Yangi environment yaratish uchun unga nom bering.

    C:\CATS_vs_DOGS> python -m venv sizning_env
  

• Virtual environmentni ishga tushiring(aktivlashtiring).

    C:\CATS_vs_DOGS> sizning_env\Scripts\activate.bat
  

• Virtual environment ichiga loyiha ishlashi uchun kerakli bo'lgan modullarni o'rnating (requirements.txt faylining ichida barchasi mavjud).

    (sizning_env) C:\CATS_vs_DOGS> pip install -r requirements.txt
  

4. Proyektni ishlatish uchun jupyter notebook ni ishga tushiring.

    (sizning_env) C:\CATS_vs_DOGS> jupyter notebook

• Cats vs Dogs.ipynb ni ishga tushiring.
• Usbu notebookda Tensorflow.org saytidagi cats_vs_dogs datasetini o'qib olish, uni train va test datalariga ajratish, datalarni size va shape larini train uchun moslash hamda normallashtirish ko'rsatilgan.
• Dataset yordamida Convolutional Neural Network ishlab chiqilgan va u yordamida model train va evaluate qilingan. Model fayl ko'rinishida saqlanadi.
• Notebook yordamida saqlangan modelni load qilish va yangi test qilish datalari yordamida bashorat qilish (predict) ko'rsatib o'tilgan.
• Modelni yuklab olish

5. Proyektni streamlit yordamida deploy qilish.

    (sizning_env) C:\CATS_vs_DOGS> streamlit run streamlit.py

• Proyekt local serverda ishga tushadi va quyidagicha ko'rinishda bo'ladi:

• Rasm faylini yuklab oling va Predict tugmachasini bosing. Model yuklab olingan tasvirni qaysi turkumga tegishli ekanligini bashorat qiladi. Bundan tashqari softmaxdan chiqqan ehtimollik natijasi ham ekranga chiqadi.