Initial commit

app.py

@@ -0,0 +1,63 @@
+# app.py
+import torch
+import torchaudio
+import numpy as np
+from transformers import Wav2Vec2Processor, HubertModel
+from sklearn.preprocessing import StandardScaler
+import gradio as gr
+
+# โหลดโมเดล HuBERT
+model_name = "facebook/hubert-large-ls960-ft"
+processor = Wav2Vec2Processor.from_pretrained(model_name)
+hubert_model = HubertModel.from_pretrained(model_name)
+hubert_model.eval()
+
+# โหลดโมเดล classifier (ต้องบันทึก model.pt และ scaler.pkl ก่อน!)
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model_cls = torch.load("model.pt", map_location=device)
+model_cls.eval()
+
+import joblib
+scaler = joblib.load("scaler.pkl")
+
+def extract_mean_embedding(wav_path):
+    waveform, sample_rate = torchaudio.load(wav_path)
+    waveform = waveform.squeeze()
+    inputs = processor(waveform, sampling_rate=sample_rate, return_tensors="pt", padding=True)
+    with torch.no_grad():
+        outputs = hubert_model(**inputs)
+    embedding = outputs.last_hidden_state.mean(dim=1).squeeze().numpy()
+    return embedding
+
+def predict_water_status(file):
+    vec = extract_mean_embedding(file).reshape(1, -1)
+    vec_scaled = scaler.transform(vec)
+    vec_tensor = torch.tensor(vec_scaled, dtype=torch.float32).to(device)
+    with torch.no_grad():
+        outputs = model_cls(vec_tensor)
+        pred = outputs.argmax(dim=1).item()
+    return "🌵 ขาดน้ำ" if pred == 0 else "💧 มีน้ำเพียงพอ"
+
+with gr.Blocks() as interface:
+    gr.Markdown("# 🌱 Plant Sound Classifier (Fine-tuned)")
+    gr.Markdown("อัปโหลดเสียงพืชเพื่อทำนายสถานะ: ขาดน้ำ หรือ มีน้ำพอ")
+    audio_input = gr.Audio(type="filepath", label="🎧 อัปโหลดเสียงพืช (.wav)")
+    output_text = gr.Textbox(label="📋 ผลการทำนาย", lines=2)
+
+    with gr.Row():
+        submit_btn = gr.Button("Submit")
+        clear_btn = gr.Button("Clear")
+
+    submit_btn.click(
+        fn=predict_water_status,
+        inputs=audio_input,
+        outputs=output_text
+    )
+
+    clear_btn.click(
+        fn=lambda: (None, ""),
+        inputs=[],
+        outputs=[audio_input, output_text]
+    )
+
+interface.launch()

model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96dc8d26134177f23d6ad2c80e671d205f1152dc420070a5d3c408edb67b8101
+size 1120116

requirements.txt

@@ -0,0 +1 @@
+torchaudio\ntorch\ntransformers\ngradio\nscikit-learn

scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fbcc0275244723f605f277bd04c49d3b616f7969bdc6066c00967fcd1e2481e
+size 25191