Ankitkumarunajaima's picture
Create app.py
a4b6e9e verified
Raw History Blame Contribute Delete
11.5 kB
import streamlit as st
import cv2
import numpy as np
from groq import Groq
from fpdf import FPDF
from datetime import datetime
import sqlite3
import pandas as pd
import plotly.graph_objects as go
import io
import os
import gc
# --- 1. SETUP & THEME (v58 Base) ---
st.set_page_config(page_title="Structural AI Audit Pro v59", layout="wide")
st.markdown("""
<style>
.stApp {
background-image: linear-gradient(rgba(255, 255, 255, 0.85), rgba(255, 255, 255, 0.85)),
url('https://img.freepik.com/free-vector/abstract-architectural-blueprint-background_52683-59424.jpg') !important;
background-size: cover !important;
background-attachment: fixed !important;
}
h1 { color: #1E3A8A !important; font-weight: 800 !important; text-align: center; }
</style>
""", unsafe_allow_html=True)
st.title("πŸ—οΈ Structural Audit AI Portal v59")
# Database Setup
conn = sqlite3.connect('structural_master_v59.db', check_same_thread=False)
c = conn.cursor()
c.execute('''CREATE TABLE IF NOT EXISTS audit_logs
(date TEXT, time TEXT, location TEXT, material TEXT, width TEXT, depth TEXT, priority TEXT, cost TEXT, details TEXT)''')
conn.commit()
GROQ_API_KEY = "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"
client = Groq(api_key=GROQ_API_KEY)
# --- 2. SIDEBAR (v58 Same) ---
with st.sidebar:
st.header("πŸ›‘οΈ Engineering Panel")
if st.button("πŸ”„ Clear System Cache", use_container_width=True):
st.cache_data.clear()
st.cache_resource.clear()
st.rerun()
with st.expander("πŸ“š IS Code Reference", expanded=True):
st.write("**IS 456:2000 Standards**")
st.write("β€’ Max Width: 0.3mm\nβ€’ Repair: Grouting > 0.5mm")
st.divider()
custom_rate = st.number_input("Base Rate per mmΒ² (Rs.)", value=10.0)
base_visit_fee = st.number_input("Base Visiting Fee (Rs.)", value=200)
calib = st.slider("Calibration", 0.01, 0.10, 0.05)
sens = st.slider("Precision", 0.05, 1.0, 0.30)
# --- 3. CORE LOGIC (With Visual 3D Depth) ---
def process_analysis(img, sensitivity, calib):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (7, 7), 0)
edges = cv2.Canny(blur, int(sensitivity * 30), 100)
contours, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
marked_img = img.copy()
h_data = np.zeros(img.shape[:2], dtype=np.uint8)
max_w_px = 0.0; total_len_px = 0.0; total_area_px = 0.0
crack_pixels = []
for cnt in contours:
x, y, wb, hb = cv2.boundingRect(cnt)
if wb > 1.5 or hb > 1.5:
w_px = min(wb, hb); l_px = max(wb, hb)
if w_px > max_w_px: max_w_px = w_px
total_len_px += l_px; total_area_px += cv2.contourArea(cnt)
cv2.drawContours(marked_img, [cnt], -1, (0, 0, 255), 2)
cv2.drawContours(h_data, [cnt], -1, (255), -1)
mask = np.zeros(gray.shape, np.uint8)
cv2.drawContours(mask, [cnt], -1, 255, -1)
pixel_values = gray[mask == 255]
crack_pixels.extend(pixel_values)
heatmap = cv2.applyColorMap(cv2.GaussianBlur(h_data, (51, 51), 0), cv2.COLORMAP_JET)
mm_w = round(max_w_px * calib, 2)
mm_l = round(total_len_px * calib, 2)
mm_area = round(total_area_px * (calib ** 2), 2)
# 3D Depth Estimation logic (from v57/v58)
estimated_depth = 0.0
if crack_pixels and mm_w > 0:
avg_crack = np.mean(crack_pixels)
avg_bg = np.mean(gray)
contrast = (avg_bg - avg_crack) / (avg_bg + 1)
estimated_depth = round(mm_w * (0.8 + contrast), 2)
else:
estimated_depth = round(mm_w * 0.6, 2)
# Clamp depth to a reasonable range
if estimated_depth < 0.1: estimated_depth = 0.1
if estimated_depth > 100: estimated_depth = 100
# --- Visual 3D Depth Effect on Marked Image ---
# Intensify blue color based on depth, simulate shadow
if mm_w > 0 and estimated_depth > 0:
depth_norm = min(estimated_depth / 50.0, 1.0) # Normalize depth for intensity (max 50mm for effect)
# Create a darker, thicker blue line for deeper cracks
for cnt in contours:
cv2.drawContours(marked_img, [cnt], -1, (int(255 * depth_norm), 0, 255), int(2 + depth_norm * 3)) # Bolder blue
# Add a slight shadow effect (darker outline)
cv2.drawContours(marked_img, [cnt], -1, (int(100 * depth_norm), 0, int(100 * depth_norm)), int(1 + depth_norm * 1), lineType=cv2.LINE_AA)
return marked_img, heatmap, mm_w, mm_l, mm_area, estimated_depth
def generate_stress_curve(mm_w, material, location):
E_base = 35000 if material == "Concrete" else 5000
loc_factor = {"Column": 1.5, "Beam": 1.3, "Slab": 1.0, "Wall": 0.8}.get(location, 0.6) # Enhanced loc_factor
E = E_base * loc_factor
# Failure strain depends on width and material (Concrete fails earlier for same strain)
if material == "Concrete":
failure_strain = 0.002 + (mm_w / 60)
else: # Plaster can deform more before 'failure' in terms of structural load
failure_strain = 0.004 + (mm_w / 50)
strain = np.linspace(0, failure_strain * 1.5, 100)
# Stress-Strain Parabolic Curve (IS 456 for concrete, adjusted for plaster)
stress = np.maximum(E * strain * (1 - (strain / (2.2 * failure_strain))), 0) # Adjusted constant for better curve shape
fig = go.Figure()
fig.add_trace(go.Scatter(x=strain, y=stress, mode='lines', name='Structural Integrity', line=dict(color='#ff5733', width=3)))
# Calculate failure point (max stress)
max_stress_idx = np.argmax(stress)
fig.add_trace(go.Scatter(x=[strain[max_stress_idx]], y=[stress[max_stress_idx]],
mode='markers', name='Failure Point',
marker=dict(color='red', size=10, symbol='x', line=dict(width=2, color='DarkRed')))) # Clearer X mark
fig.update_layout(title=f"Physics Plot: {location} ({material})",
xaxis_title="Strain (Ξ΅)", yaxis_title="Stress (MPa)",
template="plotly_dark", height=280)
return fig
def get_priority_v54(width, depth, material):
# Priority logic now more sensitive to depth
if material == "Plaster":
if width < 1.0 and depth < 5.0: return "LOW", "🟒", "Safe (Surface Plaster Crack)"
elif (1.0 <= width < 3.0 and depth < 10.0) or (depth >= 5.0 and depth < 15.0): return "MEDIUM", "🟑", "Warning"
else: return "HIGH", "πŸ”΄", "Deep Crack (Internal Check Req.)"
else: # Concrete
if width < 0.3 and depth < 10.0: return "LOW", "🟒", "Safe (Hairline)"
elif (0.3 <= width < 0.7 and depth < 20.0) or (depth >= 10.0 and depth < 30.0): return "MEDIUM", "🟑", "Warning (Structural)"
else: return "HIGH", "πŸ”΄", "CRITICAL (Immediate Action!)"
# --- 4. TABS UI ---
tab1, tab2, tab3 = st.tabs(["πŸš€ Batch Audit", "πŸ“Έ Live Cam", "πŸ“œ History"])
with tab1:
files = st.file_uploader("Upload Crack Images", accept_multiple_files=True, key="file_uploader_v59")
if files:
c1, c2 = st.columns(2)
with c1: mat = st.radio("Material:", ["Concrete", "Plaster"], horizontal=True)
with c2: loc = st.selectbox("Location:", ["Wall", "Column", "Beam", "Slab", "Surface"])
if st.button("Execute Batch Analysis", use_container_width=True):
gc.collect()
pdf = FPDF()
pdf.set_font("Arial", size=10) # Set default font
for i, f in enumerate(files):
img = cv2.imdecode(np.frombuffer(f.read(), np.uint8), 1)
m_img, h_img, w, l, area, depth = process_analysis(img, sens, calib)
priority, emoji, p_text = get_priority_v54(w, depth, mat)
total_repair = round((area * custom_rate * (1.5 if priority=="MEDIUM" else 2.5 if priority=="HIGH" else 1.0)) + base_visit_fee, 2)
# UI Preview
st.subheader(f"Result {i+1}: {f.name} ({emoji} {p_text})")
m1, m2, m3, m4 = st.columns(4)
m1.metric("Width", f"{w} mm")
m2.metric("Depth", f"{depth} mm")
m3.metric("Status", p_text)
m4.metric("Estimate", f"Rs. {total_repair}")
# Image Layout Fix: Use columns to display images side-by-side
img_col1, img_col2, img_col3 = st.columns(3)
with img_col1: st.image(img, caption="Original", use_column_width=True)
with img_col2: st.image(m_img, caption="Marked (3D Depth)", use_column_width=True) # Updated caption
with img_col3: st.image(h_img, caption="Thermal Heatmap", use_column_width=True)
graph_key = f"graph_{datetime.now().strftime('%Y%m%d%H%M%S%f')}_{i}"
fig = generate_stress_curve(w, mat, loc)
st.plotly_chart(fig, use_container_width=True, key=graph_key)
prompt = f"Analyze {w}mm width and {depth}mm depth crack on {loc} ({mat}). Priority: {priority}. Give detailed Cause & Repair solutions."
ai_resp = client.chat.completions.create(messages=[{"role":"user","content":prompt}], model="llama-3.1-8b-instant").choices[0].message.content
st.info(ai_resp)
# Database (v58 Same)
c.execute("INSERT INTO audit_logs VALUES (?,?,?,?,?,?,?,?,?)",
(datetime.now().strftime('%Y-%m-%d'), datetime.now().strftime('%H:%M:%S'), loc, mat, f"{w}mm", f"{depth}mm", priority, f"Rs. {total_repair}", ai_resp))
conn.commit()
# PDF Export (v58 Same)
pdf.add_page()
pdf.set_font("Arial", 'B', 14)
pdf.cell(0, 10, f"STRUCTURAL REPORT - {f.name}", 1, 1, 'C')
pdf.set_font("Arial", size=10)
pdf.cell(0, 8, f"Location: {loc} | Material: {mat} | Width: {w}mm | Depth: {depth}mm", ln=1)
pdf.cell(0, 8, f"Priority: {priority} | Estimate: Rs. {total_repair}", ln=1)
o_p, m_p, h_p, g_p = f"o_{i}.jpg", f"m_{i}.jpg", f"h_{i}.jpg", f"g_{i}.png"
cv2.imwrite(o_p, img); cv2.imwrite(m_p, m_img); cv2.imwrite(h_p, h_img); fig.write_image(g_p)
pdf.image(o_p, 10, 40, 90, 60); pdf.image(m_p, 105, 40, 90, 60)
pdf.image(h_p, 10, 105, 90, 60); pdf.image(g_p, 105, 105, 90, 60)
pdf.add_page()
pdf.set_font("Arial", size=10)
pdf.multi_cell(0, 6, txt=ai_resp.encode('latin-1', 'ignore').decode('latin-1'))
st.download_button("πŸ“₯ Download PDF Report", bytes(pdf.output(dest='S')), "Audit_Report.pdf", use_container_width=True)
with tab2: # Live Cam (v58 with depth info)
live = st.camera_input("Scan Crack")
if live:
img_l = cv2.imdecode(np.frombuffer(live.read(), np.uint8), 1)
m_l, _, w_l, _, _, d_l = process_analysis(img_l, sens, calib)
st.image(m_l, caption=f"Detected: {w_l}mm | Depth: {d_l}mm", use_column_width=True)
with tab3: # History (v58 Same)
history = pd.read_sql_query("SELECT * FROM audit_logs ORDER BY date DESC", conn)
st.dataframe(history, use_container_width=True)