Geographic_Analysis_Toolkit / spatial_diffusion.py
dhammo2's picture
Upload 3 files
617c74e verified
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from shapely.geometry import Point, Polygon
import random
import datetime
import gradio as gr
import tempfile
import os
import requests
import json
from typing import List, Tuple, Optional, Dict, Any, Union
def fetch_osm_exclusion_zones(bounds: Tuple[float, float, float, float], exclusion_types: List[str]) -> Optional[Any]:
"""
Fetch exclusion zones from OpenStreetMap using Overpass API.
Args:
bounds: (min_lat, min_lon, max_lat, max_lon) bounding box
exclusion_types: List of exclusion types to fetch
Returns:
GeoDataFrame with exclusion polygons or None if failed
"""
try:
import geopandas as gpd
from shapely.geometry import Polygon, MultiPolygon, LineString
# Overpass API endpoint
overpass_url = "http://overpass-api.de/api/interpreter"
# Build Overpass query based on selected exclusion types
queries = []
if "Water bodies" in exclusion_types:
# Get both water polygons AND linear waterways
queries.extend([
# Water area polygons
f'way["natural"="water"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'relation["natural"="water"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["landuse"="reservoir"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["water"="lake"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["water"="pond"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
# Linear waterways (rivers, streams, canals)
f'way["waterway"="river"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["waterway"="stream"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["waterway"="canal"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});'
])
if "Parks & green spaces" in exclusion_types:
queries.extend([
f'way["leisure"="park"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["landuse"="forest"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["landuse"="grass"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["natural"="wood"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});'
])
if "Industrial areas" in exclusion_types:
queries.extend([
f'way["landuse"="industrial"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});',
f'way["landuse"="commercial"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});'
])
if "Major roads" in exclusion_types:
queries.extend([
f'way["highway"~"motorway|trunk|primary"]({bounds[0]},{bounds[1]},{bounds[2]},{bounds[3]});'
])
if not queries:
return None
# Build complete Overpass query
overpass_query = f"""
[out:json][timeout:25];
(
{chr(10).join(queries)}
);
out geom;
"""
print(f"Fetching OSM data for exclusion zones: {exclusion_types}")
# Make request to Overpass API
response = requests.get(overpass_url, params={'data': overpass_query})
response.raise_for_status()
data = response.json()
if 'elements' not in data or not data['elements']:
print("No exclusion zones found in the specified area")
return None
# Convert OSM data to polygons
polygons = []
zone_types = []
for element in data['elements']:
try:
if element['type'] == 'way' and 'geometry' in element:
tags = element.get('tags', {})
# Determine what type of feature this is
zone_type = None
if 'natural' in tags and tags['natural'] == 'water':
zone_type = 'Water'
elif 'landuse' in tags and tags['landuse'] == 'reservoir':
zone_type = 'Water'
elif 'water' in tags:
zone_type = 'Water'
elif 'waterway' in tags and tags['waterway'] in ['river', 'stream', 'canal']:
zone_type = 'Water'
elif 'leisure' in tags and tags['leisure'] == 'park':
zone_type = 'Park'
elif 'landuse' in tags and tags['landuse'] in ['forest', 'grass']:
zone_type = 'Green space'
elif 'natural' in tags and tags['natural'] == 'wood':
zone_type = 'Forest'
elif 'landuse' in tags and tags['landuse'] in ['industrial', 'commercial']:
zone_type = 'Industrial/Commercial'
elif 'highway' in tags:
zone_type = 'Major road'
if zone_type is None:
continue
# Create polygon from way geometry
coords = [(node['lon'], node['lat']) for node in element['geometry']]
# Handle different geometry types
if 'waterway' in tags or 'highway' in tags:
# For linear features (rivers, roads), create a buffered polygon from the line
if len(coords) >= 2:
try:
line = LineString(coords)
# Buffer size depends on feature type
if 'waterway' in tags:
if tags['waterway'] == 'river':
buffer_size = 50 / 111320 # Rivers: ~50 meters
elif tags['waterway'] == 'canal':
buffer_size = 30 / 111320 # Canals: ~30 meters
else: # streams
buffer_size = 20 / 111320 # Streams: ~20 meters
else: # highways
buffer_size = 25 / 111320 # Roads: ~25 meters
polygon = line.buffer(buffer_size)
if polygon.is_valid and polygon.area > 0:
polygons.append(polygon)
zone_types.append(zone_type)
except Exception as e:
print(f"Error buffering linear feature: {str(e)}")
continue
else:
# For areas, create closed polygons
if len(coords) > 2:
# Close polygon if not already closed
if coords[0] != coords[-1]:
coords.append(coords[0])
if len(coords) >= 4: # Valid polygon needs at least 4 points
try:
polygon = Polygon(coords)
if polygon.is_valid and polygon.area > 0:
polygons.append(polygon)
zone_types.append(zone_type)
except Exception as e:
print(f"Error creating polygon: {str(e)}")
continue
except Exception as e:
print(f"Error processing OSM element: {str(e)}")
continue
if not polygons:
print("No valid polygons found in OSM data")
return None
# Create GeoDataFrame
gdf = gpd.GeoDataFrame(
{'zone_type': zone_types},
geometry=polygons,
crs='EPSG:4326'
)
print(f"Successfully fetched {len(gdf)} exclusion zones from OpenStreetMap")
print(f"Zone types found: {gdf['zone_type'].value_counts().to_dict()}")
return gdf
except ImportError:
print("GeoPandas not available for OSM processing")
return None
except requests.exceptions.RequestException as e:
print(f"Error fetching data from OpenStreetMap: {str(e)}")
return None
except Exception as e:
print(f"Error processing OpenStreetMap data: {str(e)}")
return None
def calculate_bounds_from_points(input_df: pd.DataFrame, buffer_km: float = 2.0) -> Tuple[float, float, float, float]:
"""Calculate bounding box around input points with buffer"""
# Get min/max coordinates
min_lat = input_df['lat'].min()
max_lat = input_df['lat'].max()
min_lon = input_df['lon'].min()
max_lon = input_df['lon'].max()
# Add buffer (approximate conversion from km to degrees)
buffer_deg = buffer_km / 111.0 # Rough conversion: 1 degree ≈ 111 km
return (
min_lat - buffer_deg, # min_lat
min_lon - buffer_deg, # min_lon
max_lat + buffer_deg, # max_lat
max_lon + buffer_deg # max_lon
)
class SpatialDiffuser:
"""
Class for performing spatial diffusion - takes points with counts and diffuses them
according to specified distributions within given radii, with optional exclusion zones.
"""
def __init__(self):
self.distribution_methods = {
"uniform": self._uniform_distribution,
"normal": self._normal_distribution,
"exponential_decay": self._exponential_decay,
"distance_weighted": self._distance_weighted
}
def diffuse_points(self,
input_data: pd.DataFrame,
distribution_type: str = "uniform",
global_radius: Optional[float] = None,
time_start: Optional[datetime.datetime] = None,
time_end: Optional[datetime.datetime] = None,
seed: Optional[int] = None,
exclusion_zones_gdf: Optional[Any] = None) -> pd.DataFrame:
"""
Generate diffused points based on input coordinates and counts.
Args:
input_data: DataFrame with columns: lat, lon, count, radius (optional)
distribution_type: Type of spatial distribution to use
global_radius: Radius to use for all points if not specified individually (in meters)
time_start: Start time for temporal distribution
time_end: End time for temporal distribution
seed: Random seed for reproducible results
exclusion_zones_gdf: GeoDataFrame with polygons to exclude points from
Returns:
DataFrame with columns: lat, lon, source_id, timestamp (if temporal)
"""
# Set random seed if provided
if seed is not None:
np.random.seed(seed)
random.seed(seed)
if distribution_type not in self.distribution_methods:
raise ValueError(f"Distribution type '{distribution_type}' not supported. Choose from: {list(self.distribution_methods.keys())}")
# Initialize list to hold all generated points
all_points = []
# Generate points for each input location
for idx, row in input_data.iterrows():
# Get radius (either from row or global)
radius = row.get('radius', global_radius)
if radius is None:
raise ValueError("Radius must be specified either globally or per point")
# Get count
count = int(row['count'])
if count <= 0:
continue
# Generate points with exclusion zone filtering
new_points = self._generate_points_with_exclusions(
lat=row['lat'],
lon=row['lon'],
count=count,
radius=radius,
distribution_type=distribution_type,
exclusion_zones_gdf=exclusion_zones_gdf
)
# Add source identifier
source_ids = [idx] * len(new_points)
# Add timestamps if temporal distribution is requested
if time_start is not None and time_end is not None:
timestamps = self._generate_timestamps(len(new_points), time_start, time_end)
# Combine points with metadata
for i, point in enumerate(new_points):
all_points.append({
'lat': point[0],
'lon': point[1],
'source_id': source_ids[i],
'timestamp': timestamps[i]
})
else:
# Combine points with metadata without timestamps
for i, point in enumerate(new_points):
all_points.append({
'lat': point[0],
'lon': point[1],
'source_id': source_ids[i]
})
# Convert to DataFrame
result = pd.DataFrame(all_points)
return result
def _generate_points_with_exclusions(self, lat: float, lon: float, count: int, radius: float,
distribution_type: str, exclusion_zones_gdf: Optional[Any] = None) -> List[Tuple[float, float]]:
"""Generate points while avoiding exclusion zones"""
if exclusion_zones_gdf is None or len(exclusion_zones_gdf) == 0:
# No exclusion zones, use normal generation
return self.distribution_methods[distribution_type](lat, lon, count, radius)
try:
import geopandas as gpd
from shapely.geometry import Point
valid_points = []
max_attempts = count * 10 # Generate up to 10x more points to account for exclusions
attempts = 0
# Ensure exclusion zones are in the same CRS as our points (WGS84)
if exclusion_zones_gdf.crs is None:
exclusion_zones_gdf = exclusion_zones_gdf.set_crs('EPSG:4326')
elif exclusion_zones_gdf.crs != 'EPSG:4326':
exclusion_zones_gdf = exclusion_zones_gdf.to_crs('EPSG:4326')
while len(valid_points) < count and attempts < max_attempts:
# Generate a batch of points
batch_size = min(count * 2, max_attempts - attempts)
candidate_points = self.distribution_methods[distribution_type](
lat, lon, batch_size, radius
)
# Check each point against exclusion zones
for point in candidate_points:
if len(valid_points) >= count:
break
point_geom = Point(point[1], point[0]) # lon, lat for Point
# Check if point intersects with any exclusion zone
is_excluded = False
for _, exclusion_zone in exclusion_zones_gdf.iterrows():
if point_geom.intersects(exclusion_zone.geometry):
is_excluded = True
break
if not is_excluded:
valid_points.append(point)
attempts += batch_size
# If we couldn't generate enough valid points, warn the user
if len(valid_points) < count:
print(f"Warning: Could only generate {len(valid_points)} valid points out of {count} requested for location ({lat}, {lon}). Exclusion zones may be too restrictive.")
return valid_points
except ImportError:
print("GeoPandas not available for exclusion zone processing. Generating points without exclusions.")
return self.distribution_methods[distribution_type](lat, lon, count, radius)
except Exception as e:
print(f"Error processing exclusion zones: {str(e)}. Generating points without exclusions.")
return self.distribution_methods[distribution_type](lat, lon, count, radius)
def _uniform_distribution(self, lat: float, lon: float, count: int, radius: float) -> List[Tuple[float, float]]:
"""Generate points uniformly distributed within a circle"""
points = []
for _ in range(count):
# Generate a random angle and distance
angle = random.uniform(0, 2 * np.pi)
# Uniform distribution needs square root to avoid clustering in center
r = radius * np.sqrt(random.uniform(0, 1))
# Convert polar coordinates to Cartesian
x = r * np.cos(angle)
y = r * np.sin(angle)
# Convert meters to approximate degrees (this is a simplification)
# A more accurate implementation would use proper geographic projections
lat_offset = y / 111320 # 1 degree latitude is approximately 111320 meters
# Longitude degrees vary with latitude, so adjust accordingly
lon_offset = x / (111320 * np.cos(np.radians(lat)))
new_lat = lat + lat_offset
new_lon = lon + lon_offset
points.append((new_lat, new_lon))
return points
def _normal_distribution(self, lat: float, lon: float, count: int, radius: float) -> List[Tuple[float, float]]:
"""Generate points with normal distribution (more concentrated near center)"""
points = []
# Standard deviation as a fraction of radius
std_dev = radius / 3 # 3 sigma rule - 99.7% of points within radius
for _ in range(count):
# Generate points using normal distribution
while True:
# Generate x and y offsets using normal distribution
x = np.random.normal(0, std_dev)
y = np.random.normal(0, std_dev)
# Calculate distance from center
distance = np.sqrt(x**2 + y**2)
# If point is within radius, keep it
if distance <= radius:
break
# Convert meters to approximate degrees
lat_offset = y / 111320
lon_offset = x / (111320 * np.cos(np.radians(lat)))
new_lat = lat + lat_offset
new_lon = lon + lon_offset
points.append((new_lat, new_lon))
return points
def _exponential_decay(self, lat: float, lon: float, count: int, radius: float) -> List[Tuple[float, float]]:
"""Generate points with exponential decay from center"""
points = []
# Rate parameter - controls how quickly density decreases with distance
rate = 3.0 / radius # Higher value = steeper decay
for _ in range(count):
# Generate random angle
angle = random.uniform(0, 2 * np.pi)
# Generate distance with exponential distribution
# Use rejection sampling to ensure points are within radius
while True:
# Generate exponential random variable
r = random.expovariate(rate)
if r <= radius:
break
# Convert polar coordinates to Cartesian
x = r * np.cos(angle)
y = r * np.sin(angle)
# Convert meters to approximate degrees
lat_offset = y / 111320
lon_offset = x / (111320 * np.cos(np.radians(lat)))
new_lat = lat + lat_offset
new_lon = lon + lon_offset
points.append((new_lat, new_lon))
return points
def _distance_weighted(self, lat: float, lon: float, count: int, radius: float) -> List[Tuple[float, float]]:
"""
Generate points with a custom distance-weighted distribution
(more points at medium distances than at center or edge)
"""
points = []
for _ in range(count):
# Generate random angle
angle = random.uniform(0, 2 * np.pi)
# Custom distribution - more weight at middle distances
# Generate r² with beta distribution (concentration in middle)
r_squared = random.betavariate(2, 2) # Beta with alpha=beta=2 has peak in middle
r = np.sqrt(r_squared) * radius
# Convert polar coordinates to Cartesian
x = r * np.cos(angle)
y = r * np.sin(angle)
# Convert meters to approximate degrees
lat_offset = y / 111320
lon_offset = x / (111320 * np.cos(np.radians(lat)))
new_lat = lat + lat_offset
new_lon = lon + lon_offset
points.append((new_lat, new_lon))
return points
def _generate_timestamps(self, count: int, start_time: datetime.datetime, end_time: datetime.datetime) -> List[datetime.datetime]:
"""Generate uniformly distributed timestamps"""
timestamps = []
# Convert to timestamps for easier calculations
start_ts = start_time.timestamp()
end_ts = end_time.timestamp()
for _ in range(count):
# Generate a random timestamp between start and end
random_ts = random.uniform(start_ts, end_ts)
timestamp = datetime.datetime.fromtimestamp(random_ts)
timestamps.append(timestamp)
# Sort timestamps chronologically
timestamps.sort()
return timestamps
def create_visualization(input_df, output_df, show_basemap=False, exclusion_zones_gdf=None):
"""Create visualization of input and diffused points"""
fig, ax = plt.subplots(figsize=(12, 10))
# Set background color
fig.patch.set_facecolor('white')
ax.set_facecolor('#f8f9fa')
# Define colors for different exclusion zone types
exclusion_colors = {
'Water': '#4FC3F7', # Light blue
'Park': '#66BB6A', # Green
'Green space': '#81C784', # Light green
'Forest': '#4CAF50', # Dark green
'Industrial/Commercial': '#90A4AE', # Grey
'Major road': '#FFD54F', # Yellow
'Other': '#FFAB91' # Light orange
}
# If basemap is requested, convert to Web Mercator and add basemap
if show_basemap:
try:
import contextily as ctx
import geopandas as gpd
from shapely.geometry import Point
# Create GeoDataFrames for proper projection
input_gdf = gpd.GeoDataFrame(
input_df,
geometry=[Point(lon, lat) for lon, lat in zip(input_df['lon'], input_df['lat'])],
crs='EPSG:4326'
)
output_gdf = gpd.GeoDataFrame(
output_df,
geometry=[Point(lon, lat) for lon, lat in zip(output_df['lon'], output_df['lat'])],
crs='EPSG:4326'
)
# Convert to Web Mercator for basemap compatibility
input_gdf_merc = input_gdf.to_crs('EPSG:3857')
output_gdf_merc = output_gdf.to_crs('EPSG:3857')
# Plot exclusion zones first (if provided) with color coding
if exclusion_zones_gdf is not None and len(exclusion_zones_gdf) > 0:
try:
exclusion_zones_merc = exclusion_zones_gdf.to_crs('EPSG:3857')
# Group by zone type and plot with appropriate colors
plotted_types = set()
for zone_type in exclusion_zones_merc['zone_type'].unique():
zone_subset = exclusion_zones_merc[exclusion_zones_merc['zone_type'] == zone_type]
color = exclusion_colors.get(zone_type, exclusion_colors['Other'])
# Only add label for first occurrence of each type
label = zone_type if zone_type not in plotted_types else None
if label:
plotted_types.add(zone_type)
zone_subset.plot(ax=ax, color=color, alpha=0.6, edgecolor='white',
linewidth=0.5, label=label)
except Exception as e:
print(f"Error plotting exclusion zones: {str(e)}")
# Extract coordinates for plotting
input_x = input_gdf_merc.geometry.x
input_y = input_gdf_merc.geometry.y
output_x = output_gdf_merc.geometry.x
output_y = output_gdf_merc.geometry.y
# Plot diffused points first (so they appear behind source points)
ax.scatter(output_x, output_y,
alpha=0.7, color='#FF9800', s=12, label=f'Generated Points (n={len(output_df)})',
edgecolors='white', linewidth=0.3)
# Draw radius circles first (so they appear behind everything else)
for idx, row in input_df.iterrows():
radius = row.get('radius', None)
if radius is not None:
# Convert center point to Web Mercator
center_point = gpd.GeoDataFrame(
[1], geometry=[Point(row['lon'], row['lat'])], crs='EPSG:4326'
).to_crs('EPSG:3857')
center_x = center_point.geometry.x.iloc[0]
center_y = center_point.geometry.y.iloc[0]
# Draw circle (radius is already in meters, which matches Web Mercator units)
circle = plt.Circle((center_x, center_y), radius,
fill=False, color='#9C27B0', linestyle='--',
alpha=0.5, linewidth=2)
ax.add_patch(circle)
# Plot source points with circles sized by count
min_size = 100
max_size = 800
if len(input_df) > 1:
size_range = input_df['count'].max() - input_df['count'].min()
if size_range > 0:
sizes = min_size + (input_df['count'] - input_df['count'].min()) / size_range * (max_size - min_size)
else:
sizes = [min_size] * len(input_df)
else:
sizes = [max_size]
# Plot source points in purple
ax.scatter(input_x, input_y,
s=sizes, c='#9C27B0', alpha=0.9,
edgecolors='white', linewidth=2,
label='Source Points (size = count)', zorder=5)
# Add count labels next to source points
for idx, row in input_df.iterrows():
point_merc = gpd.GeoDataFrame(
[1], geometry=[Point(row['lon'], row['lat'])], crs='EPSG:4326'
).to_crs('EPSG:3857')
x_merc = point_merc.geometry.x.iloc[0]
y_merc = point_merc.geometry.y.iloc[0]
ax.annotate(f'{int(row["count"])}',
(x_merc, y_merc),
xytext=(8, 8), textcoords='offset points',
fontsize=10, fontweight='bold', color='white',
bbox=dict(boxstyle='round,pad=0.3', facecolor='#9C27B0', alpha=0.8),
zorder=6)
# Add basemap
try:
ctx.add_basemap(ax, crs='EPSG:3857', source=ctx.providers.CartoDB.Positron, alpha=0.8)
basemap_added = True
except Exception as e:
print(f"Could not add basemap: {str(e)}")
basemap_added = False
# Set axis labels for Web Mercator
ax.set_xlabel('Easting (Web Mercator)', fontsize=12)
ax.set_ylabel('Northing (Web Mercator)', fontsize=12)
# Use projected coordinates for limits
x_coords = list(input_x) + list(output_x)
y_coords = list(input_y) + list(output_y)
except ImportError:
print("Contextily not available for basemap. Falling back to simple plot.")
show_basemap = False
except Exception as e:
print(f"Error creating basemap: {str(e)}. Falling back to simple plot.")
show_basemap = False
# Fallback to simple plot if basemap fails or is not requested
if not show_basemap:
# Plot exclusion zones first (if provided) with color coding
if exclusion_zones_gdf is not None and len(exclusion_zones_gdf) > 0:
try:
# Ensure exclusion zones are in WGS84
if exclusion_zones_gdf.crs != 'EPSG:4326':
exclusion_zones_gdf = exclusion_zones_gdf.to_crs('EPSG:4326')
# Plot zones by type with appropriate colors
plotted_types = set()
for idx, zone in exclusion_zones_gdf.iterrows():
zone_type = zone.get('zone_type', 'Other')
color = exclusion_colors.get(zone_type, exclusion_colors['Other'])
# Only add label for first occurrence of each type
label = zone_type if zone_type not in plotted_types else None
if label:
plotted_types.add(zone_type)
if zone.geometry.geom_type == 'Polygon':
x, y = zone.geometry.exterior.xy
ax.fill(x, y, color=color, alpha=0.6, edgecolor='white',
linewidth=0.5, label=label)
elif zone.geometry.geom_type == 'MultiPolygon':
for poly in zone.geometry.geoms:
x, y = poly.exterior.xy
ax.fill(x, y, color=color, alpha=0.6, edgecolor='white',
linewidth=0.5, label=label)
label = None # Only label the first polygon
except Exception as e:
print(f"Error plotting exclusion zones: {str(e)}")
# Plot diffused points first (so they appear behind source points) - orange
ax.scatter(output_df['lon'], output_df['lat'],
alpha=0.7, color='#FF9800', s=12, label=f'Generated Points (n={len(output_df)})',
edgecolors='white', linewidth=0.3)
# Draw radius circles first (so they appear behind everything else) - purple
for idx, row in input_df.iterrows():
radius = row.get('radius', None)
if radius is not None:
# Approximate conversion from meters to degrees
radius_deg_lat = radius / 111320
radius_deg_lon = radius / (111320 * np.cos(np.radians(row['lat'])))
# Use the average as an approximation
radius_deg = (radius_deg_lat + radius_deg_lon) / 2
# Draw circle in purple
circle = plt.Circle((row['lon'], row['lat']), radius_deg,
fill=False, color='#9C27B0', linestyle='--',
alpha=0.5, linewidth=2)
ax.add_patch(circle)
# Plot source points with circles sized by count - purple
min_size = 100
max_size = 800
if len(input_df) > 1:
size_range = input_df['count'].max() - input_df['count'].min()
if size_range > 0:
sizes = min_size + (input_df['count'] - input_df['count'].min()) / size_range * (max_size - min_size)
else:
sizes = [min_size] * len(input_df)
else:
sizes = [max_size]
# Plot source points in purple
ax.scatter(input_df['lon'], input_df['lat'],
s=sizes, c='#9C27B0', alpha=0.9,
edgecolors='white', linewidth=2,
label='Source Points (size = count)', zorder=5)
# Add count labels next to source points with purple background
for idx, row in input_df.iterrows():
ax.annotate(f'{int(row["count"])}',
(row['lon'], row['lat']),
xytext=(8, 8), textcoords='offset points',
fontsize=10, fontweight='bold', color='white',
bbox=dict(boxstyle='round,pad=0.3', facecolor='#9C27B0', alpha=0.8),
zorder=6)
# Standard coordinate labels
ax.set_xlabel('Longitude', fontsize=12)
ax.set_ylabel('Latitude', fontsize=12)
# Use original coordinates for limits
x_coords = list(input_df['lon']) + list(output_df['lon'])
y_coords = list(input_df['lat']) + list(output_df['lat'])
# Common styling
title = 'Spatial Diffusion Results'
if show_basemap:
title += ' (with Basemap)'
if exclusion_zones_gdf is not None and len(exclusion_zones_gdf) > 0:
title += ' - Exclusion Zones Applied'
subtitle = 'Purple source points sized by count, orange generated points, dashed circles show diffusion radius'
ax.set_title(f'{title}\n{subtitle}',
fontsize=14, fontweight='bold', pad=20)
# Legend with better positioning
legend = ax.legend(loc='upper right', bbox_to_anchor=(1, 1),
frameon=True, fancybox=True, shadow=True)
legend.get_frame().set_facecolor('white')
legend.get_frame().set_alpha(0.9)
# Add grid (lighter for basemap)
grid_alpha = 0.2 if show_basemap else 0.3
ax.grid(True, alpha=grid_alpha, linestyle='-', linewidth=0.5)
# Make equal aspect ratio
ax.set_aspect('equal', 'box')
# Add some padding around the data
x_margin = (max(x_coords) - min(x_coords)) * 0.1
y_margin = (max(y_coords) - min(y_coords)) * 0.1
if x_margin == 0: # Handle case where all points have same x-coordinate
x_margin = 1000 if show_basemap else 0.01
if y_margin == 0: # Handle case where all points have same y-coordinate
y_margin = 1000 if show_basemap else 0.01
ax.set_xlim(min(x_coords) - x_margin, max(x_coords) + x_margin)
ax.set_ylim(min(y_coords) - y_margin, max(y_coords) + y_margin)
# Tight layout
plt.tight_layout()
return fig
def process_csv(file_obj, distribution_type, global_radius, show_basemap, auto_exclusions, exclusion_file, include_time, time_start, time_end, seed):
"""Process input CSV and generate diffused points"""
try:
# Read input CSV
df = pd.read_csv(file_obj.name)
# Validate required columns
required_cols = ['lat', 'lon', 'count']
if not all(col in df.columns for col in required_cols):
return None, f"Error: CSV must contain columns: {', '.join(required_cols)}"
# Convert global_radius to float if provided
if global_radius and global_radius.strip():
try:
global_radius = float(global_radius)
except ValueError:
return None, "Error: Global radius must be a number"
else:
global_radius = None
# If global radius not provided, check for radius column
if 'radius' not in df.columns:
return None, "Error: Either provide a global radius or include a 'radius' column in the CSV"
# Convert seed to int if provided
if seed and seed.strip():
try:
seed = int(seed)
except ValueError:
return None, "Error: Seed must be an integer"
else:
seed = None
# Process exclusion zones
exclusion_zones_gdf = None
# First, try manual file upload (takes priority)
if exclusion_file is not None:
try:
import geopandas as gpd
# Determine file type and read accordingly
file_extension = os.path.splitext(exclusion_file.name)[1].lower()
if file_extension in ['.geojson', '.json']:
exclusion_zones_gdf = gpd.read_file(exclusion_file.name)
elif file_extension == '.gpkg':
exclusion_zones_gdf = gpd.read_file(exclusion_file.name)
elif file_extension == '.shp':
exclusion_zones_gdf = gpd.read_file(exclusion_file.name)
else:
return None, f"Error: Unsupported exclusion zone file format: {file_extension}"
# Ensure CRS is set
if exclusion_zones_gdf.crs is None:
exclusion_zones_gdf = exclusion_zones_gdf.set_crs('EPSG:4326')
print(f"Loaded {len(exclusion_zones_gdf)} custom exclusion zones from {exclusion_file.name}")
except ImportError:
return None, "Error: GeoPandas required for exclusion zones processing"
except Exception as e:
return None, f"Error reading exclusion zones file: {str(e)}"
# If no manual file, try automatic exclusions from OpenStreetMap
elif auto_exclusions and len(auto_exclusions) > 0:
try:
# Calculate bounds around input points
bounds = calculate_bounds_from_points(df)
print(f"Fetching automatic exclusions for bounds: {bounds}")
# Fetch OSM data
exclusion_zones_gdf = fetch_osm_exclusion_zones(bounds, auto_exclusions)
if exclusion_zones_gdf is not None:
print(f"Fetched {len(exclusion_zones_gdf)} exclusion zones from OpenStreetMap")
else:
print("No exclusion zones found in OpenStreetMap for this area")
except Exception as e:
print(f"Warning: Could not fetch automatic exclusions: {str(e)}")
# Continue without exclusions rather than failing completely
exclusion_zones_gdf = None
# Process time if requested
if include_time:
if not time_start or not time_end:
return None, "Error: If time distribution is enabled, both start and end times must be provided"
try:
time_start_dt = datetime.datetime.strptime(time_start, "%Y-%m-%d %H:%M:%S")
time_end_dt = datetime.datetime.strptime(time_end, "%Y-%m-%d %H:%M:%S")
if time_start_dt >= time_end_dt:
return None, "Error: End time must be after start time"
except ValueError:
return None, "Error: Invalid time format. Use YYYY-MM-DD HH:MM:SS"
else:
time_start_dt = None
time_end_dt = None
# Create diffuser and generate diffused points
diffuser = SpatialDiffuser()
result_df = diffuser.diffuse_points(
input_data=df,
distribution_type=distribution_type,
global_radius=global_radius,
time_start=time_start_dt,
time_end=time_end_dt,
seed=seed,
exclusion_zones_gdf=exclusion_zones_gdf
)
# Create temporary CSV file
temp_file = "diffused_points.csv"
result_df.to_csv(temp_file, index=False)
# Create visualization with basemap and exclusion zones
fig = create_visualization(df, result_df, show_basemap, exclusion_zones_gdf)
return fig, temp_file
except Exception as e:
return None, f"Error: {str(e)}"
def create_diffusion_interface():
"""Create Gradio interface for the spatial diffusion tool"""
with gr.Blocks() as diffusion_interface:
gr.Markdown("## 🗺️ Spatial Diffusion Tool")
with gr.Row():
with gr.Column(scale=1):
# Move description into the left column for better space usage
gr.Markdown("""
### About This Tool
Transform aggregated geographic points with counts into individual points using spatial diffusion methods.
**Input CSV Format:**
- `lat`: Latitude of source point
- `lon`: Longitude of source point
- `count`: Number of points to generate
- `radius`: (Optional) Diffusion radius in meters
**Distribution Types:**
- **Uniform**: Equal probability throughout circle
- **Normal**: Higher density near center
- **Exponential Decay**: Density decreases from center
- **Distance-Weighted**: More points at medium distances
""")
# Input controls
input_file = gr.File(label="Input CSV File", file_types=[".csv"])
# Distribution options grouped together
gr.Markdown("### 🎯 Distribution Options")
with gr.Row():
distribution = gr.Dropdown(
choices=["uniform", "normal", "exponential_decay", "distance_weighted"],
value="uniform",
label="Distribution Type",
scale=2
)
seed = gr.Textbox(
label="Random Seed (optional)",
placeholder="e.g. 42",
scale=1
)
global_radius = gr.Textbox(
label="Global Radius (meters)",
placeholder="Only if radius column not in CSV"
)
# Temporal controls in distribution section
with gr.Accordion("⏰ Temporal Distribution (Optional)", open=False):
include_time = gr.Checkbox(label="Enable Temporal Distribution", value=False)
with gr.Group() as time_group:
time_start = gr.Textbox(
label="Start Time",
placeholder="YYYY-MM-DD HH:MM:SS"
)
time_end = gr.Textbox(
label="End Time",
placeholder="YYYY-MM-DD HH:MM:SS"
)
# Map and exclusion options grouped together
gr.Markdown("### 🗺️ Map & Exclusion Options")
show_basemap = gr.Checkbox(
label="Show underlying map (requires internet)",
value=False
)
gr.Markdown("*Adds geographic context with street/satellite imagery*")
# Automatic exclusion zones - no default selection
auto_exclusions = gr.CheckboxGroup(
label="Auto-exclude from OpenStreetMap:",
choices=["Water bodies", "Parks & green spaces", "Industrial areas", "Major roads"],
value=[] # No default selections
)
# Advanced manual exclusion zones
with gr.Accordion("🔧 Advanced: Custom Exclusion Zones", open=False):
exclusion_file = gr.File(
label="Upload custom shapefile (optional)",
file_types=[".geojson", ".json", ".gpkg", ".shp"]
)
gr.Markdown("*Overrides automatic exclusions if provided*")
process_btn = gr.Button(
"🎯 Generate Diffused Points",
variant="primary",
size="lg"
)
with gr.Column(scale=2):
# Give more space to visualization
plot_output = gr.Plot(
label="📍 Spatial Diffusion Visualization",
show_label=True
)
with gr.Row():
with gr.Column(scale=2):
file_output = gr.File(label="📥 Download Generated Points")
with gr.Column(scale=1):
gr.Markdown(
"""
**Legend:**
🟣 Source points (sized by count)
🟠 Generated points
⭕ Diffusion radius
🟦 Water bodies
🟢 Parks & green spaces
⬜ Industrial areas
🟡 Major roads
"""
)
# Set up event handlers
process_btn.click(
fn=process_csv,
inputs=[input_file, distribution, global_radius, show_basemap, auto_exclusions, exclusion_file, include_time, time_start, time_end, seed],
outputs=[plot_output, file_output]
)
# Show/hide time inputs based on checkbox
include_time.change(
fn=lambda x: gr.update(visible=x),
inputs=[include_time],
outputs=[time_group]
)
return diffusion_interface
if __name__ == "__main__":
# This allows the module to be run directly for testing
app = create_diffusion_interface()
app.launch()