jonas/snow_trail
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

stylized 1-bit rendering

Browse Source
This commit is contained in:
Jonas H
2026-01-21 11:04:55 +01:00
parent 5d2eca0393
commit 2422106725
40 changed files with 2859 additions and 366 deletions
Download Patch File Download Diff File Expand all files Collapse all files

594
blender/scripts/generate_flowmap.py Normal file
View File

@@ -0,0 +1,594 @@
import bpy
import bmesh
import numpy as np
from pathlib import Path
from mathutils import Vector
from multiprocessing import Pool, cpu_count
def simple_blur(data, iterations=1):
"""
Simple 3x3 box blur for 2D vector fields.
"""
result = data.copy()
h, w, c = data.shape
for _ in range(iterations):
blurred = np.zeros_like(result)
for y in range(h):
for x in range(w):
count = 0
total = np.zeros(c)
for dy in [-1, 0, 1]:
for dx in [-1, 0, 1]:
ny, nx = y + dy, x + dx
if 0 <= ny < h and 0 <= nx < w:
total += result[ny, nx]
count += 1
blurred[y, x] = total / count if count > 0 else result[y, x]
result = blurred
return result
def sample_curve_points(curve_obj, samples_per_segment=50):
"""
Sample points along a Blender curve object (Bezier, NURBS, etc).
Args:
curve_obj: Blender curve object
samples_per_segment: Number of samples per curve segment
Returns:
List of Vector world-space positions
"""
curve_obj.data.resolution_u = samples_per_segment
depsgraph = bpy.context.evaluated_depsgraph_get()
curve_eval = curve_obj.evaluated_get(depsgraph)
mesh_from_curve = curve_eval.to_mesh()
points = [curve_obj.matrix_world @ v.co for v in mesh_from_curve.vertices]
curve_eval.to_mesh_clear()
return points
def compute_arc_lengths(points):
"""
Compute cumulative arc length at each point along a curve.
Args:
points: List of Vector positions
Returns:
List of floats representing cumulative distance from curve start
"""
if len(points) == 0:
return []
arc_lengths = [0.0]
cumulative = 0.0
for i in range(1, len(points)):
segment_length = (points[i] - points[i-1]).length
cumulative += segment_length
arc_lengths.append(cumulative)
return arc_lengths
def get_path_curves(collection_name="Paths"):
"""
Get all curve objects from a collection, or all curves with 'path' in the name.
Args:
collection_name: Name of collection containing path curves
Returns:
List of curve objects
"""
collection = bpy.data.collections.get(collection_name)
if collection:
curves = [obj for obj in collection.objects if obj.type == 'CURVE']
print(f"Found {len(curves)} curves in collection '{collection_name}'")
return curves
print(f"Collection '{collection_name}' not found. Searching for curves with 'path' in name...")
curves = [obj for obj in bpy.data.objects if obj.type == 'CURVE' and 'path' in obj.name.lower()]
if not curves:
print("No path curves found. Looking for any curve objects...")
curves = [obj for obj in bpy.data.objects if obj.type == 'CURVE']
return curves
def closest_point_on_segment_2d(point, seg_start, seg_end):
"""
Find closest point on a 2D line segment to a given point.
Args:
point: Vector2 (x, y)
seg_start: Vector2 segment start
seg_end: Vector2 segment end
Returns:
Vector2 closest point on segment
"""
segment = seg_end - seg_start
point_vec = point - seg_start
segment_len_sq = segment.length_squared
if segment_len_sq < 1e-8:
return seg_start
t = max(0.0, min(1.0, point_vec.dot(segment) / segment_len_sq))
return seg_start + segment * t
def process_row_chunk(args):
"""
Process a chunk of rows for parallel computation.
Args:
args: Tuple of (start_row, end_row, width, height_total, min_x, max_x, min_y, max_y, path_segments_simple)
Returns:
Tuple of (flow_field_chunk, distance_field_chunk, arc_length_field_chunk, segment_id_chunk)
"""
start_row, end_row, width, height_total, min_x, max_x, min_y, max_y, path_segments_simple = args
height = end_row - start_row
flow_field_chunk = np.zeros((height, width, 2), dtype=np.float32)
distance_field_chunk = np.zeros((height, width), dtype=np.float32)
arc_length_field_chunk = np.zeros((height, width), dtype=np.float32)
segment_id_chunk = np.zeros((height, width), dtype=np.int32)
for local_y in range(height):
y = start_row + local_y
for x in range(width):
u = x / width
v = y / height_total
world_x = min_x + u * (max_x - min_x)
world_y = min_y + v * (max_y - min_y)
world_pos_2d = np.array([world_x, world_y])
min_dist = float('inf')
closest_point = None
closest_arc_length = 0.0
closest_segment_id = -1
for seg_id, (seg_start, seg_end, arc_start, arc_end) in enumerate(path_segments_simple):
segment = seg_end - seg_start
point_vec = world_pos_2d - seg_start
segment_len_sq = np.dot(segment, segment)
if segment_len_sq < 1e-8:
closest_on_segment = seg_start
t = 0.0
else:
t = max(0.0, min(1.0, np.dot(point_vec, segment) / segment_len_sq))
closest_on_segment = seg_start + segment * t
dist = np.linalg.norm(closest_on_segment - world_pos_2d)
if dist < min_dist:
min_dist = dist
closest_point = closest_on_segment
closest_arc_length = arc_start + t * (arc_end - arc_start)
closest_segment_id = seg_id
if closest_point is not None:
direction = closest_point - world_pos_2d
flow_field_chunk[local_y, x, 0] = direction[0]
flow_field_chunk[local_y, x, 1] = -direction[1]
distance_field_chunk[local_y, x] = min_dist
arc_length_field_chunk[local_y, x] = closest_arc_length
segment_id_chunk[local_y, x] = closest_segment_id
return (start_row, flow_field_chunk, distance_field_chunk, arc_length_field_chunk, segment_id_chunk)
def generate_flowmap_from_paths(terrain_obj, path_curves, resolution=1024, blur_iterations=2):
"""
Generate a flowmap texture showing direction toward nearest path.
Args:
terrain_obj: Blender mesh object (the terrain plane)
path_curves: List of Blender curve objects representing paths
resolution: Output texture resolution (square)
blur_iterations: Number of smoothing passes for flow field
Returns:
Blender image containing encoded flowmap
"""
print(f"Generating flowmap from {len(path_curves)} path curves")
print(f"Output resolution: {resolution}×{resolution}")
mesh = terrain_obj.data
bm = bmesh.new()
bm.from_mesh(mesh)
bm.verts.ensure_lookup_table()
world_matrix = terrain_obj.matrix_world
verts_world = [world_matrix @ v.co for v in bm.verts]
if len(verts_world) == 0:
raise ValueError("Terrain mesh has no vertices")
xs = [v.x for v in verts_world]
ys = [v.y for v in verts_world]
min_x, max_x = min(xs), max(xs)
min_y, max_y = min(ys), max(ys)
print(f"Terrain bounds: X=[{min_x:.2f}, {max_x:.2f}], Y=[{min_y:.2f}, {max_y:.2f}]")
print("Sampling path curves...")
all_path_points = []
for curve_obj in path_curves:
points = sample_curve_points(curve_obj, samples_per_segment=50)
all_path_points.extend(points)
print(f" {curve_obj.name}: {len(points)} points")
print(f"Total path points: {len(all_path_points)}")
if len(all_path_points) == 0:
raise ValueError("No path points sampled. Check that path curves exist and have geometry.")
print("Building line segments from path points with arc lengths...")
path_segments = []
current_segment_start = 0
for curve_obj in path_curves:
curve_points = sample_curve_points(curve_obj, samples_per_segment=50)
arc_lengths = compute_arc_lengths(curve_points)
for i in range(len(curve_points) - 1):
path_segments.append((curve_points[i], curve_points[i + 1], arc_lengths[i], arc_lengths[i + 1]))
print(f"Created {len(path_segments)} line segments with arc length data")
print("Generating flow field toward nearest path...")
h, w = resolution, resolution
flow_field = np.zeros((h, w, 2), dtype=np.float32)
distance_field = np.zeros((h, w), dtype=np.float32)
arc_length_field = np.zeros((h, w), dtype=np.float32)
segment_id_field = np.zeros((h, w), dtype=np.int32)
path_segments_simple = [
(np.array([seg_start.x, seg_start.y]), np.array([seg_end.x, seg_end.y]), arc_start, arc_end)
for seg_start, seg_end, arc_start, arc_end in path_segments
]
num_cores = cpu_count()
chunk_size = max(1, h // num_cores)
print(f"Using {num_cores} CPU cores with chunk size {chunk_size} rows")
chunks = []
for start_row in range(0, h, chunk_size):
end_row = min(start_row + chunk_size, h)
chunks.append((start_row, end_row, w, h, min_x, max_x, min_y, max_y, path_segments_simple))
with Pool(processes=num_cores) as pool:
total_chunks = len(chunks)
print(f"Processing {total_chunks} chunks...")
results = []
for i, result in enumerate(pool.imap(process_row_chunk, chunks)):
results.append(result)
progress = (i + 1) / total_chunks * 100
print(f" Progress: {i+1}/{total_chunks} chunks ({progress:.1f}%)")
print("Assembling results from parallel workers...")
for start_row, flow_chunk, distance_chunk, arc_length_chunk, segment_id_chunk in results:
end_row = start_row + flow_chunk.shape[0]
flow_field[start_row:end_row, :, :] = flow_chunk
distance_field[start_row:end_row, :] = distance_chunk
arc_length_field[start_row:end_row, :] = arc_length_chunk
segment_id_field[start_row:end_row, :] = segment_id_chunk
print(f"Distance range: {distance_field.min():.2f} to {distance_field.max():.2f}")
print(f"Arc length range: {arc_length_field.min():.2f} to {arc_length_field.max():.2f}")
print("Normalizing flow vectors...")
magnitudes = np.sqrt(flow_field[:, :, 0]**2 + flow_field[:, :, 1]**2)
magnitudes = np.maximum(magnitudes, 1e-8)
flow_field[:, :, 0] /= magnitudes
flow_field[:, :, 1] /= magnitudes
if blur_iterations > 0:
print(f"Applying distance-based blur (max {blur_iterations} iterations)...")
original_flow = flow_field.copy()
print(" Creating maximally blurred version...")
blurred_flow = simple_blur(flow_field, iterations=blur_iterations)
magnitudes = np.sqrt(blurred_flow[:, :, 0]**2 + blurred_flow[:, :, 1]**2)
magnitudes = np.maximum(magnitudes, 1e-8)
blurred_flow[:, :, 0] /= magnitudes
blurred_flow[:, :, 1] /= magnitudes
print(" Blending based on distance to path...")
max_distance = 60.0
distance_normalized = np.clip(distance_field / max_distance, 0.0, 1.0)
blend_factor = distance_normalized[:, :, np.newaxis]
flow_field = original_flow * (1.0 - blend_factor) + blurred_flow * blend_factor
magnitudes = np.sqrt(flow_field[:, :, 0]**2 + flow_field[:, :, 1]**2)
magnitudes = np.maximum(magnitudes, 1e-8)
flow_field[:, :, 0] /= magnitudes
flow_field[:, :, 1] /= magnitudes
print(" Distance-based blur complete!")
print("Saving segment ID debug image...")
num_segments = len(path_segments)
segment_colors = np.random.RandomState(42).rand(num_segments, 3).astype(np.float32)
segment_img = np.zeros((resolution, resolution, 4), dtype=np.float32)
for y in range(resolution):
for x in range(resolution):
seg_id = segment_id_field[y, x]
if seg_id >= 0 and seg_id < num_segments:
segment_img[y, x, 0:3] = segment_colors[seg_id]
segment_img[y, x, 3] = 1.0
segment_debug_img = bpy.data.images.new(
"Segment_Debug", width=resolution, height=resolution, alpha=True, float_buffer=True
)
segment_debug_img.pixels[:] = segment_img.flatten()
segment_debug_img.filepath_raw = str(
Path(bpy.data.filepath).parent.parent / "textures" / "path_segment_debug.png"
)
segment_debug_img.file_format = 'PNG'
segment_debug_img.save()
bpy.data.images.remove(segment_debug_img)
print(f" Saved to textures/path_segment_debug.png ({num_segments} segments)")
print("Saving distance field debug image...")
distance_normalized = np.clip(distance_field / 50.0, 0.0, 1.0)
distance_img = np.zeros((resolution, resolution, 4), dtype=np.float32)
distance_img[:, :, 0] = distance_normalized
distance_img[:, :, 1] = distance_normalized
distance_img[:, :, 2] = distance_normalized
distance_img[:, :, 3] = 1.0
debug_img = bpy.data.images.new(
"Distance_Debug", width=resolution, height=resolution, alpha=True, float_buffer=True
)
debug_img.pixels[:] = distance_img.flatten()
debug_img.filepath_raw = str(
Path(bpy.data.filepath).parent.parent / "textures" / "path_distance_debug.png"
)
debug_img.file_format = 'PNG'
debug_img.save()
bpy.data.images.remove(debug_img)
print(" Saved to textures/path_distance_debug.png")
print("Saving direction field debug image...")
direction_img = np.zeros((resolution, resolution, 4), dtype=np.float32)
cardinal_dirs = np.array([
[0.0, 1.0],
[0.707, 0.707],
[1.0, 0.0],
[0.707, -0.707],
[0.0, -1.0],
[-0.707, -0.707],
[-1.0, 0.0],
[-0.707, 0.707],
])
cardinal_colors = np.array([
[76, 120, 168],
[242, 142, 43],
[225, 87, 89],
[118, 183, 178],
[89, 161, 79],
[237, 201, 72],
[176, 122, 161],
[255, 157, 167],
]) / 255.0
flow_flat = flow_field.reshape(-1, 2)
dots = flow_flat @ cardinal_dirs.T
closest_dir_indices = np.argmax(dots, axis=1)
direction_img[:, :, 0:3] = cardinal_colors[closest_dir_indices].reshape(resolution, resolution, 3)
direction_img[:, :, 3] = 1.0
print("Drawing compass in lower right corner...")
compass_radius = resolution // 8
compass_center_x = resolution - compass_radius - 20
compass_center_y = compass_radius + 20
for y in range(resolution):
for x in range(resolution):
dx = x - compass_center_x
dy = y - compass_center_y
dist = np.sqrt(dx * dx + dy * dy)
if dist <= compass_radius:
angle = np.arctan2(dy, dx)
sector = int(((angle + np.pi / 2) % (2 * np.pi)) / (np.pi / 4)) % 8
direction_img[y, x, 0:3] = cardinal_colors[sector]
direction_debug_img = bpy.data.images.new(
"Direction_Debug", width=resolution, height=resolution, alpha=True, float_buffer=True
)
direction_debug_img.pixels[:] = direction_img.flatten()
direction_debug_img.filepath_raw = str(
Path(bpy.data.filepath).parent.parent / "textures" / "path_direction_debug.png"
)
direction_debug_img.file_format = 'PNG'
direction_debug_img.save()
bpy.data.images.remove(direction_debug_img)
print(" Saved to textures/path_direction_debug.png (8-color cardinal directions)")
print("\n=== DEBUG: Sample flow field values ===")
print(f"Terrain bounds: X=[{min_x}, {max_x}], Y=[{min_y}, {max_y}]")
sample_pixels = [
(resolution // 4, resolution // 4),
(resolution // 2, resolution // 2),
(3 * resolution // 4, 3 * resolution // 4),
]
for py, px in sample_pixels:
u = px / resolution
v = py / resolution
world_x = min_x + u * (max_x - min_x)
world_y = min_y + v * (max_y - min_y)
flow_x = flow_field[py, px, 0]
flow_y = flow_field[py, px, 1]
dots = [np.dot([flow_x, flow_y], cardinal_dirs[i]) for i in range(8)]
best_dir = np.argmax(dots)
dir_names = ["North", "NE", "East", "SE", "South", "SW", "West", "NW"]
print(f" Pixel [{px}, {py}] -> World [{world_x:.1f}, {world_y:.1f}]")
print(f" Flow: [{flow_x:.3f}, {flow_y:.3f}] -> {dir_names[best_dir]}")
print(f" Encoded: R={flow_x * 0.5 + 0.5:.3f}, G={flow_y * 0.5 + 0.5:.3f}")
print("Encoding to RGBA texture...")
flow_encoded_x = flow_field[:, :, 0] * 0.5 + 0.5
flow_encoded_y = flow_field[:, :, 1] * 0.5 + 0.5
distance_normalized = np.clip(distance_field / max_distance, 0.0, 1.0)
arc_length_repeat = 100.0
arc_length_normalized = np.fmod(arc_length_field, arc_length_repeat) / arc_length_repeat
print(f"Distance encoding: 0.0 = on path, 1.0 = {max_distance}+ units away")
print(f"Arc length encoding: repeating pattern every {arc_length_repeat} world units")
flowmap = np.zeros((resolution, resolution, 4), dtype=np.float32)
flowmap[:, :, 0] = flow_encoded_x
flowmap[:, :, 1] = flow_encoded_y
flowmap[:, :, 2] = distance_normalized
flowmap[:, :, 3] = arc_length_normalized
print(f"Creating Blender image...")
output_img = bpy.data.images.new(
name="Path_Flowmap",
width=resolution,
height=resolution,
alpha=True,
float_buffer=True
)
output_img.pixels[:] = flowmap.flatten()
bm.free()
return output_img
def save_flowmap(output_path, resolution=1024, blur_iterations=3, terrain_name="TerrainPlane", path_collection="Paths"):
"""
Main function to generate and save flowmap from paths in current Blender file.
Args:
output_path: Path to save PNG flowmap
resolution: Output texture resolution
blur_iterations: Smoothing iterations
terrain_name: Name of terrain object
path_collection: Name of collection or search pattern for path curves
"""
terrain_obj = bpy.data.objects.get(terrain_name)
if not terrain_obj:
print(f"Object '{terrain_name}' not found. Searching for terrain-like objects...")
for obj in bpy.data.objects:
if obj.type == 'MESH':
print(f" Found mesh: {obj.name}")
if 'terrain' in obj.name.lower() or 'plane' in obj.name.lower():
terrain_obj = obj
print(f" -> Using: {obj.name}")
break
if not terrain_obj:
raise ValueError(
f"Object '{terrain_name}' not found and no terrain mesh detected. "
f"Available objects: {[obj.name for obj in bpy.data.objects if obj.type == 'MESH']}"
)
print(f"Using terrain object: {terrain_obj.name}")
if terrain_obj.type != 'MESH':
raise ValueError(f"Object '{terrain_obj.name}' is not a mesh")
path_curves = get_path_curves(path_collection)
if not path_curves:
raise ValueError(
f"No path curves found. Create curves and add them to a '{path_collection}' collection "
f"or name them with 'path' in the name. Available curves: "
f"{[obj.name for obj in bpy.data.objects if obj.type == 'CURVE']}"
)
print(f"Found {len(path_curves)} path curves:")
for curve in path_curves:
print(f" - {curve.name}")
flowmap_img = generate_flowmap_from_paths(
terrain_obj,
path_curves,
resolution=resolution,
blur_iterations=blur_iterations
)
flowmap_img.filepath_raw = str(output_path)
flowmap_img.file_format = 'OPEN_EXR'
flowmap_img.save()
bpy.data.images.remove(flowmap_img)
print(f"\n{'='*60}")
print(f"Flowmap generation complete!")
print(f"Output: {output_path}")
print(f"Resolution: {resolution}×{resolution}")
print(f"Paths used: {len(path_curves)}")
print(f"Format: OpenEXR (32-bit float)")
print(f"{'='*60}")
print("\nChannel encoding:")
print(" R channel: X direction toward nearest path (0.5=none, <0.5=left, >0.5=right)")
print(" G channel: Y direction toward nearest path (0.5=none, <0.5=down, >0.5=up)")
print(" B channel: Distance to nearest path (0.0=on path, 1.0=far away)")
print(" A channel: Arc length along path (0.0-1.0, repeating every 100 units)")
print("\nTo use in shader:")
print(" vec4 flowmap_sample = texture(flowmap, uv);")
print(" vec2 direction = flowmap_sample.rg * 2.0 - 1.0; // Direction to path")
print(" float distance = flowmap_sample.b; // Distance to path")
print(" float arc_length = flowmap_sample.a * 100.0; // Arc length in world units")
if __name__ == "__main__":
project_root = Path(bpy.data.filepath).parent.parent
output_path = project_root / "textures" / "terrain_flowmap.exr"
output_path.parent.mkdir(parents=True, exist_ok=True)
save_flowmap(
output_path=output_path,
resolution=1024,
blur_iterations=5,
terrain_name="TerrainPlane",
path_collection="Paths"
)