Create app.py

app.py

@@ -0,0 +1,109 @@
+import gradio as gr
+import pandas as pd
+import matplotlib.pyplot as plt
+import random
+from datetime import datetime, timedelta
+from transformers import pipeline, AutoProcessor, Qwen2VLForConditionalGeneration, Blip2ForConditionalGeneration
+from PIL import Image
+import torch
+
+# Function to simulate sensor data
+def generate_sensor_data(num_entries):
+    data = []
+    timestamp = datetime.now()
+    for _ in range(num_entries):
+        entry = {
+            "timestamp": timestamp.strftime("%Y-%m-%d %H:%M:%S"),
+            "engine_temp": random.randint(80, 110),
+            "oil_pressure": random.randint(15, 50),
+            "tire_pressure": random.randint(29, 35),
+            "battery_voltage": round(random.uniform(11.0, 13.0), 1)
+        }
+        # Introduce anomalies
+        if random.random() < 0.1:
+            entry["engine_temp"] = random.randint(105, 120)
+        if random.random() < 0.1:
+            entry["oil_pressure"] = random.randint(10, 20)
+        if random.random() < 0.05:
+            entry["battery_voltage"] = round(random.uniform(10.5, 11.5), 1)
+        data.append(entry)
+        timestamp += timedelta(minutes=5)
+    return pd.DataFrame(data)
+
+# Generate sensor data
+sensor_data = generate_sensor_data(10)
+
+# Plot sensor data
+def plot_sensor_data(df):
+    timestamps = pd.to_datetime(df["timestamp"])
+    fig, axs = plt.subplots(2, 2, figsize=(14, 10))
+    fig.suptitle('Sensor Data', fontsize=16)
+
+    thresholds = {"engine_temp": 100, "oil_pressure": 25, "tire_pressure": 28, "battery_voltage": 11.5}
+
+    axs[0, 0].plot(timestamps, df["engine_temp"], color='red')
+    axs[0, 0].axhline(y=thresholds["engine_temp"], color='green', linestyle='--')
+    axs[0, 0].set_title("Engine Temperature (°C)")
+
+    axs[0, 1].plot(timestamps, df["oil_pressure"], color='blue')
+    axs[0, 1].axhline(y=thresholds["oil_pressure"], color='orange', linestyle='--')
+    axs[0, 1].set_title("Oil Pressure (psi)")
+
+    axs[1, 0].plot(timestamps, df["tire_pressure"], color='green')
+    axs[1, 0].axhline(y=thresholds["tire_pressure"], color='purple', linestyle='--')
+    axs[1, 0].set_title("Tire Pressure (psi)")
+
+    axs[1, 1].plot(timestamps, df["battery_voltage"], color='black')
+    axs[1, 1].axhline(y=thresholds["battery_voltage"], color='brown', linestyle='--')
+    axs[1, 1].set_title("Battery Voltage (V)")
+
+    plt.tight_layout()
+    fig_path = "sensor_plot.png"
+    plt.savefig(fig_path)
+    return fig_path
+
+# Initialize models
+# def load_models():
+#     vqa_model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-flan-t5-xl")
+#     damage_model = Qwen2VLForConditionalGeneration.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
+#     processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
+#     return vqa_model, damage_model, processor
+
+# vqa_model, damage_model, processor = load_models()
+
+# Generate recommendations
+def analyze_data(image, plot_path):
+    # Damage analysis
+    if image:
+        messages = [{"role": "user", "content": [{"type": "image", "image": image}, {"type": "text", "text": "Describe the damage."}]}]
+        text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        inputs = processor(text=[text], images=[image], padding=True, return_tensors="pt")
+        inputs = inputs.to("cuda" if torch.cuda.is_available() else "cpu")
+        output_ids = damage_model.generate(**inputs, max_new_tokens=128)
+        damage_output = processor.batch_decode(output_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+    else:
+        damage_output = "No image uploaded."
+
+    # Graph analysis
+    vqa_result = vqa_model(image=plot_path, question="What anomalies or patterns can be observed?")
+    graph_analysis = vqa_result[0]["answer"]
+
+    # Recommendations
+    recommendations = f"Recommendations based on analysis:\n\n1. {graph_analysis}\n\n2. {damage_output}"
+    return recommendations, plot_path, graph_analysis
+
+# Gradio UI
+with gr.Blocks(css=".output-text { font-family: 'Arial'; color: #222; font-size: 1rem; }") as app:
+    gr.Markdown("# 🚗 Car Health Report Generation using Generative AI")
+    with gr.Row():
+        car_image = gr.Image(type="pil", label="Upload Car Part Damage Image")
+    with gr.Row():
+        display_graph = gr.Image(plot_sensor_data(sensor_data), label="Sensor Data Over Time")
+        recommendations = gr.Textbox(label="Analysis & Recommendations", placeholder="Insights will appear here...")
+        graph_insights = gr.Textbox(label="Graph Insights", placeholder="Graph insights will appear here...")
+    data_table = gr.Dataframe(sensor_data, label="Generated Sensor Data (Table View)", row_count=(10, "fixed"), interactive=False)
+
+    # Realistic colors and UI layout for a polished look
+    car_image.change(fn=analyze_data, inputs=[car_image, display_graph], outputs=[recommendations, display_graph, graph_insights])
+
+app.launch()