Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+final_model.keras filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,146 @@
+
+import gradio as gr
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+import tensorflow as tf
+from tensorflow.keras.models import load_model
+from scipy.signal import butter, filtfilt
+import pywt
+import tempfile
+import os
+
+# Load model
+model = load_model("final_model.keras")
+
+label_map = {0: "Bradycardia", 1: "Tachycardia", 2: "VFib", 3: "VTach", 4: "Normal"}
+color_map = {
+    "Bradycardia": "blue",
+    "Tachycardia": "orange",
+    "VFib": "red",
+    "VTach": "purple",
+    "Normal": "green"
+}
+
+def preprocess_signal(sig, fs):
+    b, a = butter(4, [0.5, 8], btype='bandpass', fs=fs)
+    filtered = filtfilt(b, a, sig)
+    smoothed = np.convolve(filtered, np.ones(5)/5, mode='same')
+    coeffs = pywt.wavedec(smoothed, 'db4', level=4)
+    coeffs[-1] = np.zeros_like(coeffs[-1])
+    cleaned = pywt.waverec(coeffs, 'db4')
+    norm = (cleaned - np.median(cleaned)) / (np.max(cleaned) - np.min(cleaned) + 1e-8)
+    return norm
+
+def segment_signal(ppg, ecg, fs, win_sec=20, overlap_sec=10):
+    win_len = fs * win_sec
+    stride = fs * overlap_sec
+    segments = []
+    for start in range(0, len(ppg) - win_len + 1, stride):
+        end = start + win_len
+        seg_ppg = preprocess_signal(ppg[start:end], fs)
+        seg_ecg = preprocess_signal(ecg[start:end], fs)
+        if len(seg_ppg) >= 2500 and len(seg_ecg) >= 2500:
+            segments.append((seg_ppg[:2500], seg_ecg[:2500]))
+    return segments
+
+def analyze_file(file, show_plot):
+    try:
+        df = pd.read_csv(file.name)
+        df.columns = df.columns.str.strip().str.upper()
+        ecg_col = next((c for c in ["ECG", "II", "III", "AVF", "V", "I"] if c in df.columns), None)
+        ppg_col = next((c for c in ["PPG", "PLETH"] if c in df.columns), None)
+        if not ecg_col or not ppg_col:
+            return f"❌ ECG or PPG column missing in {file.name}", None, None, None
+
+        fs = 250
+        if "TIME" in df.columns:
+            diffs = df["TIME"].diff().dropna().values
+            if len(diffs) > 0:
+                fs = int(round(1 / np.median(diffs)))
+
+        ecg = df[ecg_col].values
+        ppg = df[ppg_col].values
+
+        segments = segment_signal(ppg, ecg, fs)
+        if not segments:
+            return f"❌ Insufficient signal duration in {file.name}", None, None, None
+
+        ppg_input = np.array([s[0] for s in segments])[:, :, np.newaxis]
+        ecg_input = np.array([s[1] for s in segments])[:, :, np.newaxis]
+        preds = model.predict([ppg_input, ecg_input], verbose=0)
+        pred_classes = np.argmax(preds, axis=1)
+        counts = pd.Series(pred_classes).value_counts()
+        majority_class = counts.idxmax()
+        confidence = round(100 * counts.max() / len(pred_classes), 2)
+        final_label = label_map[majority_class]
+        color = color_map[final_label]
+
+        summary = (
+            f"✅ <b>File:</b> <code>{os.path.basename(file.name)}</code><br>"
+            f"🔍 <b>Prediction:</b> <b><span style='color:{color};'>{final_label}</span></b><br>"
+            f"📊 <b>Confidence:</b> {confidence}%<br>"
+            f"🧪 <b>Segments Used:</b> {len(pred_classes)}"
+        )
+
+        segment_df = pd.DataFrame({
+            "Segment #": list(range(1, len(pred_classes)+1)),
+            "Predicted Class": [label_map[c] for c in pred_classes],
+            "Confidence": preds.max(axis=1)
+        })
+        segment_df.insert(0, "File", os.path.basename(file.name))
+
+        fig = None
+        if show_plot:
+            fig, ax = plt.subplots(2, 1, figsize=(12, 5), sharex=True)
+            ax[0].plot(ppg[:fs*10], color='blue')
+            ax[0].set_title(f"PPG Signal (First 10s) - {os.path.basename(file.name)}")
+            ax[1].plot(ecg[:fs*10], color='red')
+            ax[1].set_title(f"ECG Signal (First 10s) - {os.path.basename(file.name)}")
+            plt.tight_layout()
+
+        return summary, segment_df, fig
+
+    except Exception as e:
+        return f"❌ Error in {file.name}: {e}", None, None, None
+
+def batch_predict(files, show_plot):
+    results = []
+    all_segments = []
+    all_image_paths = []
+
+    for file in files:
+        summary, df_segment, fig = analyze_file(file, show_plot)
+        results.append(summary + "<hr>")
+        if isinstance(df_segment, pd.DataFrame):
+            all_segments.append(df_segment)
+        if fig:
+            tmp_path = tempfile.NamedTemporaryFile(delete=False, suffix=".png").name
+            fig.savefig(tmp_path)
+            plt.close(fig)
+            all_image_paths.append(tmp_path)
+
+    final_df = pd.concat(all_segments, ignore_index=True) if all_segments else None
+    temp_csv = None
+    if final_df is not None:
+        temp_csv = tempfile.NamedTemporaryFile(delete=False, suffix=".csv")
+        final_df.to_csv(temp_csv.name, index=False)
+
+    return "\n\n---\n".join(results), final_df, temp_csv.name if temp_csv else None, all_image_paths if show_plot else None
+
+gr.Interface(
+    fn=batch_predict,
+    inputs=[
+        gr.File(file_types=[".csv"], file_count="multiple", label="Upload ECG + PPG CSV(s)"),
+        gr.Checkbox(label="Show signal plot preview (first 10 seconds)", value=True)
+    ],
+    outputs=[
+        gr.HTML(label="🧠 Final Prediction Summary"),
+        gr.Dataframe(label="📄 Segment-wise Predictions"),
+        gr.File(label="⬇️ Download Segment Report"),
+        gr.Gallery(label="📈 Signal Plots (Optional)", show_label=True, visible=True)
+    ],
+    title="🩺 Arrhythmia Detection using ECG + PPG",
+    description="Upload CSV files with Time, PPG/Pleth, ECG (or II/III/AVF). Model: CNN+LSTM",
+    allow_flagging="never"
+).launch()

final_model.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:884214b83040ded0d50f9fcbbae64f2404f4a9476478480272fe4fc603ee6b9d
+size 205201368

requirements.txt

@@ -0,0 +1,7 @@
+gradio
+pandas
+numpy
+matplotlib
+tensorflow
+pywt
+scipy