Fashion-MNIST minangka set data gambar artikel Zalando, sing kalebu set latihan 60,000 conto lan set tes 10,000 conto. Saben conto minangka gambar grayscale 28 × 28, digandhengake karo label saka 10 kelas. Dataset kasebut minangka panggantos langsung kanggo set data MNIST asli kanggo benchmarking algoritma pembelajaran mesin, nyedhiyakake alternatif sing luwih tantangan amarga kerumitan lan variasi ing gambar sing gegandhengan karo fashion.
Kanggo praktis nggunakake dataset Fashion-MNIST ing Google Cloud's AI Platform, siji kudu ngetutake sawetara langkah terstruktur sing nyakup persiapan data, pelatihan model, penyebaran, lan evaluasi. Saben tahapan kasebut mbutuhake pangerten lengkap babagan dataset lan lingkungan Google Cloud.
Langkah 1: Nggawe Google Cloud Environment
Sadurunge nggunakake dataset, priksa manawa sampeyan duwe akun Google Cloud. Nggawe proyek anyar ing Google Cloud Console. Aktifake tagihan kanggo proyek sampeyan lan aktifake Cloud AI Platform API. Persiyapan iki penting amarga ngidini sampeyan nggunakake infrastruktur Google sing kuat kanggo tugas machine learning.
1. Nggawe Proyek Awan Google: Navigasi menyang Google Cloud Console lan gawe proyek anyar. Temtokake jeneng unik kanggo proyek sampeyan supaya gampang diidentifikasi.
2. Ngaktifake API: Pindhah menyang dasbor API & Layanan lan aktifake Cloud AI Platform API. API iki penting kanggo nggunakake model pembelajaran mesin ing Google Cloud.
3. Instal Cloud SDK: Ngundhuh lan nginstal Google Cloud SDK ing mesin lokal. SDK iki nyedhiyakake alat baris perintah `gcloud`, sing perlu kanggo sesambungan karo sumber daya Google Cloud sampeyan.
Langkah 2: Nyiapake Dataset Fashion-MNIST
Dataset Fashion-MNIST bisa diakses saka macem-macem sumber, kalebu repositori GitHub resmi. Penting kanggo preprocess set data kanggo mesthekake yen ana ing format sing bener kanggo model latihan ing Google Cloud.
1. Ngundhuh Dataset: Dataset kasedhiya ing macem-macem format, kalebu CSV lan NumPy arrays. Kanggo pangguna TensorFlow, bisa langsung dimuat nggunakake modul `tensorflow.keras.datasets`.
python from tensorflow.keras.datasets import fashion_mnist (train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data()
2. Data Preprocessing: Normalake nilai piksel gambar menyang kisaran [0, 1] kanthi dibagi 255. Langkah iki penting kanggo mesthekake yen model konvergen kanthi efisien sajrone latihan.
python train_images = train_images/255.0 test_images = test_images/255.0
3. Reshape lan nambah Data: Gumantung ing arsitektur model, sampeyan bisa uga kudu reshape data. Kajaba iku, nimbang teknik nggedhekake data kayata rotasi, zoom, lan flip horisontal kanggo nambah kakuwatan model.
Langkah 3: Pangembangan Model
Gawe model pembelajaran mesin sing cocog kanggo dataset Fashion-MNIST. Convolutional Neural Networks (CNNs) minangka pilihan populer amarga khasiat ing tugas klasifikasi gambar.
1. Netepake Arsitektur Model: Gunakake TensorFlow utawa PyTorch kanggo nemtokake model CNN. Arsitèktur khas bisa uga kalebu pirang-pirang lapisan konvolusi sing diikuti lapisan gabungan maksimal, lan lapisan padhet sing disambungake kanthi lengkap.
python model = tf.keras.Sequential([ tf.keras.layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)), tf.keras.layers.MaxPooling2D((2, 2)), tf.keras.layers.Conv2D(64, (3, 3), activation='relu'), tf.keras.layers.MaxPooling2D((2, 2)), tf.keras.layers.Flatten(), tf.keras.layers.Dense(64, activation='relu'), tf.keras.layers.Dense(10, activation='softmax') ])
2. Kompilasi Model: Pilih pangoptimal, fungsi mundhut, lan metrik sing cocog. Kanggo klasifikasi multi-kelas, `sparse_categorical_crossentropy` biasane digunakake.
python model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
3. Nglatih Model: Cocokake model ing data latihan. Gunakake data validasi kanggo ngawasi kinerja model lan supaya ora overfitting.
python model.fit(train_images, train_labels, epochs=10, validation_data=(test_images, test_labels))
Langkah 4: Nggunakake Model ing Google Cloud AI Platform
Sawise model dilatih, langkah sabanjure yaiku nyebarake ing Google Cloud AI Platform kanggo prediksi sing bisa diukur.
1. Simpen Model: Ekspor model sing dilatih menyang format sing kompatibel karo Google Cloud, kayata TensorFlow SavedModel.
python model.save('fashion_mnist_model')
2. Unggah Model menyang Google Cloud Storage: Gunakake alat baris perintah `gsutil` kanggo ngunggah model menyang ember Google Cloud Storage.
bash gsutil cp -r fashion_mnist_model gs://your-bucket-name/
3. Nggawe Model ing Platform AI: Ing Google Cloud Console, navigasi menyang AI Platform > Models lan gawe model anyar. Nemtokake jeneng model lan wilayah.
4. Pasang Versi Model: Nggawe versi anyar saka model kanthi nemtokake path Cloud Storage saka SavedModel. Konfigurasi jinis mesin lan pilihan skala adhedhasar kabutuhan prediksi sampeyan.
5. Nguji Panyebaran: Gunakake layanan prediksi Platform AI kanggo nyoba model sing disebarake. Sampeyan bisa ngirim panjalukan HTTP kanthi data gambar menyang titik pungkasan model lan nampa prediksi.
python from google.cloud import aiplatform project = 'your-project-id' endpoint_id = 'your-endpoint-id' location = 'us-central1' aiplatform.init(project=project, location=location) endpoint = aiplatform.Endpoint(endpoint_id=endpoint_id) # Example prediction response = endpoint.predict(instances=[test_images[0].tolist()]) print(response)
Langkah 5: Evaluasi lan Iterasi Model
Sawise nyebarake, penting kanggo ngevaluasi kinerja model lan ngulang desain kanggo nambah akurasi lan efisiensi.
1. Monitor Kinerja Model: Gunakake alat ngawasi Google Cloud kanggo nglacak metrik kinerja model kayata latensi, throughput, lan akurasi prediksi. Data iki penting banget kanggo ngenali kemacetan lan wilayah kanggo perbaikan.
2. A/B Testing: Nindakake tes A/B kanggo mbandhingake versi model sing beda. Pendekatan iki mbantu ngerteni pengaruh owah-owahan lan milih model sing paling apik.
3. Integrasi lan Penyebaran Terus-terusan (CI/CD): Ngleksanakake praktik CI/CD kanggo ngotomatisasi penyebaran versi model anyar. Persiyapan iki njamin manawa dandan dikirim kanthi cepet menyang produksi.
4. Loop Komentar: Nggawe loop umpan balik karo pangguna pungkasan kanggo ngumpulake wawasan babagan prediksi model. Gunakake umpan balik iki kanggo nyempurnakake model lan nambah relevansi karo aplikasi ing donya nyata.
5. Latihan maneh karo Data Anyar: Ajeg nganyari model karo data anyar kanggo njaga akurasi saka wektu. Praktek iki penting banget ing industri fashion, ing ngendi tren lan gaya berkembang kanthi cepet.
Dataset Fashion-MNIST nyedhiyakake kasus panggunaan praktis kanggo nyebarake model klasifikasi gambar ing Platform AI Google Cloud. Kanthi ngetutake langkah-langkah sing digarisake, siji bisa nggunakake infrastruktur Google kanthi efektif kanggo mbangun, nyebarake, lan njaga model pembelajaran mesin sing bisa diukur. Proses iki ora mung nambah akurasi lan kinerja model nanging uga njamin ditrapake kanggo skenario nyata ing industri fashion.
Google kerep nganyari Platform AI (ing taun 2024 berkembang dadi Vertex AI Platform). Yen sampeyan nemoni masalah karo nganyari iki, sampeyan uga bisa nyoba kode ing ngisor iki:
python import google.auth import google.auth.transport.requests import requests import json from tensorflow.keras.datasets import fashion_mnist import numpy as np # Load and preprocess Fashion MNIST data (train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data() test_images = test_images/255.0 # Add channel dimension to all test images test_images = test_images.reshape(-1, 28, 28, 1) # Prepare your model and project details project_id = 'project_id' model_name = 'modelname' model_version = 'V1' region = 'europe-west3' # AI Platform prediction endpoint URL url = f'https://{region}-ml.googleapis.com/v1/projects/{project_id}/models/{model_name}/versions/{model_version}:predict' # Authenticate and get the auth token credentials, _ = google.auth.default() auth_req = google.auth.transport.requests.Request() credentials.refresh(auth_req) auth_token = credentials.token # Set up headers for the request headers = { 'Authorization': f'Bearer {auth_token}', 'Content-Type': 'application/json' } class_labels = [ "T-shirt/top", "Trouser", "Pullover", "Dress", "Coat", "Sandal", "Shirt", "Sneaker", "Bag", "Ankle boot" ] # Loop through the first 6 test images for i in range(6): # Prepare the instance for prediction instance = test_images[i].tolist() # Make the request body data = json.dumps({"instances": [instance]}) # Send the request response = requests.post(url, headers=headers, data=data) response_json = response.json() # Extract the predictions predicted_probs = response_json['predictions'][0] # Get the index of the highest probability predicted_index = np.argmax(predicted_probs) predicted_label = class_labels[predicted_index] predicted_probability = predicted_probs[predicted_index] # Print the result in a more readable format print(response_json) print(f"Image {i + 1}: Predicted class: {predicted_label} ({predicted_index}) with probability {predicted_probability:.10f}")
Pitakonan lan jawaban anyar liyane babagan Sinau Mesin Cloud Google EITC/AI/GCML:
- Apa hyperparameter sing digunakake ing machine learning?
- Apa basa pemrograman kanggo sinau mesin yaiku Just Python
- Kepiye cara sinau mesin ditrapake ing jagad ilmu?
- Kepiye carane sampeyan nemtokake algoritma pembelajaran mesin sing bakal digunakake lan kepiye sampeyan nemokake?
- Apa bedane antarane Federated Learning, Edge Computing lan On-Device Machine Learning?
- Kepiye nyiyapake lan ngresiki data sadurunge latihan?
- Apa tugas lan aktivitas awal tartamtu ing proyek pembelajaran mesin?
- Apa aturan jempol kanggo nggunakake strategi lan model machine learning tartamtu?
- Parameter endi sing nuduhake yen wektune kanggo ngalih saka model linear menyang sinau jero?
- Versi Python sing paling apik kanggo nginstal TensorFlow supaya ora ana masalah sing ora ana distribusi TF?
Deleng pitakonan lan jawaban liyane ing EITC/AI/GCML Google Cloud Machine Learning