Spaces:
Sleeping
Sleeping
| import streamlit as st | |
| import torch | |
| import json | |
| import requests | |
| import os | |
| from torchvision import models, transforms | |
| from PIL import Image | |
| from urllib.request import urlretrieve | |
| # --- ATUR PATH MODEL DAN LABEL (gunakan direktori yang dapat ditulis di Hugging Face Spaces) --- | |
| BASE_DIR = "/tmp/streamlit_app" | |
| # Pastikan STREAMLIT_HOME berada di direktori yang dapat ditulis | |
| os.environ["STREAMLIT_HOME"] = BASE_DIR | |
| MODEL_DIR = os.path.join(BASE_DIR, "models") | |
| LABELS_DIR = os.path.join(BASE_DIR, "labels") | |
| os.makedirs(MODEL_DIR, exist_ok=True) | |
| os.makedirs(LABELS_DIR, exist_ok=True) | |
| MODEL_FILENAME = os.getenv("MODEL_FILENAME","mobilenetv2.pth") | |
| LABELS_FILENAME = os.getenv("LABELS_FILENAME", "labels.json") | |
| model_path = os.path.join(MODEL_DIR, MODEL_FILENAME) | |
| labels_path = os.path.join(LABELS_DIR, LABELS_FILENAME) | |
| MODEL_URL = os.getenv("MODEL_URL","https://download.pytorch.org/models/mobilenet_v2-b0353104.pth") | |
| LABELS_URL = os.getenv("LABELS_URL", "https://raw.githubusercontent.com/anishathalye/imagenet-simple-labels/master/imagenet-simple-labels.json") | |
| # --- KONFIGURASI APLIKASI --- | |
| st.set_page_config( | |
| page_title="Klasifikasi Gambar (PyTorch) 📸", | |
| page_icon="🖼️", | |
| layout="centered" | |
| ) | |
| # --- FUNGSI-FUNGSI --- | |
| def load_model(): | |
| """Memuat model MobileNetV2 dari file lokal atau mengunduh jika belum ada.""" | |
| if not os.path.exists(model_path): | |
| # st.info("Mengunduh model MobileNetV2...") | |
| try: | |
| urlretrieve(MODEL_URL, model_path) | |
| # st.success("Model berhasil diunduh.") | |
| except Exception as e: | |
| st.error(f"Gagal mengunduh model: {str(e)}") | |
| return None | |
| try: | |
| # Buat model tanpa weight | |
| model = models.mobilenet_v2(weights=None) | |
| # Muat state_dict dari file lokal | |
| state_dict = torch.load(model_path, map_location=torch.device('cpu')) | |
| model.load_state_dict(state_dict) | |
| model.eval() | |
| return model | |
| except Exception as e: | |
| st.error(f"Gagal memuat model: {str(e)}") | |
| return None | |
| def load_labels(): | |
| """Memuat label dari file lokal atau mengunduh jika belum ada.""" | |
| if not os.path.exists(labels_path): | |
| # st.info("Mengunduh label ImageNet...") | |
| try: | |
| response = requests.get(LABELS_URL) | |
| response.raise_for_status() | |
| with open(labels_path, 'w') as f: | |
| json.dump(response.json(), f) | |
| # st.success("Label berhasil diunduh.") | |
| except Exception as e: | |
| st.error(f"Gagal mengunduh label: {str(e)}") | |
| return None | |
| try: | |
| with open(labels_path, 'r') as f: | |
| labels = json.load(f) | |
| return labels | |
| except Exception as e: | |
| st.error(f"Gagal memuat label: {str(e)}") | |
| return None | |
| def preprocess_image(image): | |
| """Melakukan pra-pemrosesan gambar agar sesuai dengan input model PyTorch.""" | |
| try: | |
| # Definisikan transformasi | |
| preprocess = transforms.Compose([ | |
| transforms.Resize(256), | |
| transforms.CenterCrop(224), | |
| transforms.ToTensor(), | |
| transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), | |
| ]) | |
| # Terapkan transformasi dan tambahkan dimensi batch | |
| img_t = preprocess(image) | |
| batch_t = torch.unsqueeze(img_t, 0) | |
| return batch_t | |
| except Exception as e: | |
| st.error(f"Gagal memproses gambar: {str(e)}") | |
| return None | |
| def predict(image, model, labels): | |
| """Melakukan prediksi klasifikasi pada gambar.""" | |
| try: | |
| st.info("🧠 Model sedang menganalisis gambar...") | |
| # Pra-pemrosesan gambar | |
| batch_t = preprocess_image(image) | |
| if batch_t is None: | |
| return None | |
| # Lakukan prediksi tanpa menghitung gradien | |
| with torch.no_grad(): | |
| output = model(batch_t) | |
| # Dapatkan probabilitas dengan softmax | |
| probabilities = torch.nn.functional.softmax(output[0], dim=0) | |
| # Dapatkan 3 kelas dengan probabilitas tertinggi | |
| top3_prob, top3_catid = torch.topk(probabilities, 3) | |
| # Siapkan hasil | |
| results = [] | |
| for i in range(top3_prob.size(0)): | |
| class_name = labels[top3_catid[i]] | |
| probability = top3_prob[i].item() | |
| results.append((class_name, probability)) | |
| return results | |
| except Exception as e: | |
| st.error(f"Gagal melakukan prediksi: {str(e)}") | |
| return None | |
| # --- TAMPILAN UTAMA APLIKASI --- | |
| st.title("🖼️ Aplikasi Klasifikasi Gambar (PyTorch)") | |
| st.write( | |
| "Unggah sebuah gambar, dan AI akan mencoba menebak objek apa yang ada di dalamnya! " | |
| "Aplikasi ini menggunakan model **MobileNetV2** dari PyTorch." | |
| ) | |
| # Muat model dan label | |
| try: | |
| model = load_model() | |
| labels = load_labels() | |
| if model is None or labels is None: | |
| st.error("Aplikasi tidak dapat dijalankan karena gagal memuat model atau label.") | |
| st.stop() | |
| except Exception as e: | |
| st.error(f"Kesalahan saat inisialisasi aplikasi: {str(e)}") | |
| st.stop() | |
| # Komponen untuk unggah file | |
| uploaded_file = st.file_uploader( | |
| "Pilih sebuah gambar...", | |
| type=["jpg", "jpeg", "png"], | |
| help="Format file yang didukung: JPG, JPEG, PNG" | |
| ) | |
| if uploaded_file is not None: | |
| try: | |
| # Buka dan tampilkan gambar yang diunggah | |
| image = Image.open(uploaded_file).convert('RGB') | |
| st.image(image, caption='Gambar yang Anda Unggah', use_column_width=True) | |
| # Tombol untuk memulai klasifikasi | |
| if st.button('✨ Klasifikasikan Gambar Ini!'): | |
| with st.spinner('Tunggu sebentar...'): | |
| # Lakukan prediksi | |
| predictions = predict(image, model, labels) | |
| if predictions is not None: | |
| st.subheader("✅ Hasil Prediksi Teratas:") | |
| for i, (label, score) in enumerate(predictions): | |
| st.write(f"{i+1}. **{label.replace('_', ' ').title()}** - Keyakinan: {score:.2%}") | |
| else: | |
| st.error("Prediksi gagal. Silakan coba lagi atau unggah gambar lain.") | |
| except Exception as e: | |
| st.error(f"Kesalahan saat memproses gambar yang diunggah: {str(e)}") | |
| # Tambahan debugging untuk membantu identifikasi | |
| st.write("Detail error: Periksa koneksi internet atau format gambar.") | |
| st.divider() | |
| # st.markdown( | |
| # "Dibuat dengan ❤️ menggunakan [Streamlit](https://streamlit.io), [PyTorch](https://pytorch.org/) & [Hugging Face Spaces](https://huggingface.co/spaces)." | |
| # ) | |