autoforge 1.9.2.2__tar.gz → 1.9.3__tar.gz

{autoforge-1.9.2.2 → autoforge-1.9.3}/PKG-INFO

@@ -1,13 +1,13 @@
 Metadata-Version: 2.4
 Name: AutoForge
-Version: 1.9.2.2
+Version: 1.9.3
 Summary: AutoForge is a Python tool for generating 3D printed layered models from an input image.
 Author-email: Hendric Voss <hvoss@techfak.uni-bielefeld.de>
 License: CC BY-NC-SA 4.0
-Requires-Python: >=3.9
+Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: numpy
-Requires-Dist: torch>=2.6.0
+Requires-Dist: torch>=2.9.1
 Requires-Dist: torchvision>=0.21.0
 Requires-Dist: transformers>=4.48.1
 Requires-Dist: opencv-python>=4.2.0.32
@@ -158,6 +158,7 @@ This will generate separate STL files for each color, allowing you to print face
 - `--pruning_max_layer` Max number of layers allowed after pruning (default: 75).
 - `--random_seed` Random seed for reproducibility (default: 0 (disabled)).
 - `--mps` Flag to use the Metal Performance Shaders (MPS) backend if available.
+- `--no-spike-removal` Disable spike removal for the final STL (not recommended).
 
 - `--tensorboard` Flag to enable TensorBoard logging.
 - `--run_name` *(Optional)* Name of the run used for TensorBoard logging.

{autoforge-1.9.2.2 → autoforge-1.9.3}/README.md

@@ -133,6 +133,7 @@ This will generate separate STL files for each color, allowing you to print face
 - `--pruning_max_layer` Max number of layers allowed after pruning (default: 75).
 - `--random_seed` Random seed for reproducibility (default: 0 (disabled)).
 - `--mps` Flag to use the Metal Performance Shaders (MPS) backend if available.
+- `--no-spike-removal` Disable spike removal for the final STL (not recommended).
 
 - `--tensorboard` Flag to enable TensorBoard logging.
 - `--run_name` *(Optional)* Name of the run used for TensorBoard logging.

{autoforge-1.9.2.2 → autoforge-1.9.3}/pyproject.toml

@@ -4,15 +4,15 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "AutoForge"
-version="1.9.2.2"
+version="1.9.3"
 description = "AutoForge is a Python tool for generating 3D printed layered models from an input image."
 authors = [{ name = "Hendric Voss", email = "hvoss@techfak.uni-bielefeld.de" }]
 readme = "README.md"
-requires-python = ">=3.9"
+requires-python = ">=3.10"
 license = { text = "CC BY-NC-SA 4.0" }
 dependencies = [
   "numpy",
-  "torch>=2.6.0",
+  "torch>=2.9.1",
   "torchvision>=0.21.0",
   "transformers>=4.48.1",
   "opencv-python>=4.2.0.32",

{autoforge-1.9.2.2 → autoforge-1.9.3}/src/AutoForge.egg-info/PKG-INFO

@@ -1,13 +1,13 @@
 Metadata-Version: 2.4
 Name: AutoForge
-Version: 1.9.2.2
+Version: 1.9.3
 Summary: AutoForge is a Python tool for generating 3D printed layered models from an input image.
 Author-email: Hendric Voss <hvoss@techfak.uni-bielefeld.de>
 License: CC BY-NC-SA 4.0
-Requires-Python: >=3.9
+Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 Requires-Dist: numpy
-Requires-Dist: torch>=2.6.0
+Requires-Dist: torch>=2.9.1
 Requires-Dist: torchvision>=0.21.0
 Requires-Dist: transformers>=4.48.1
 Requires-Dist: opencv-python>=4.2.0.32
@@ -158,6 +158,7 @@ This will generate separate STL files for each color, allowing you to print face
 - `--pruning_max_layer` Max number of layers allowed after pruning (default: 75).
 - `--random_seed` Random seed for reproducibility (default: 0 (disabled)).
 - `--mps` Flag to use the Metal Performance Shaders (MPS) backend if available.
+- `--no-spike-removal` Disable spike removal for the final STL (not recommended).
 
 - `--tensorboard` Flag to enable TensorBoard logging.
 - `--run_name` *(Optional)* Name of the run used for TensorBoard logging.

{autoforge-1.9.2.2 → autoforge-1.9.3}/src/AutoForge.egg-info/SOURCES.txt

@@ -53,4 +53,5 @@ tests/test_output_helper.py
 tests/test_output_helper_extended.py
 tests/test_perception_and_pruning.py
 tests/test_property_based.py
-tests/test_pruning_helper.py
+tests/test_pruning_helper.py
+tests/test_remove_height_spikes.py

{autoforge-1.9.2.2 → autoforge-1.9.3}/src/AutoForge.egg-info/requires.txt

@@ -1,5 +1,5 @@
 numpy
-torch>=2.6.0
+torch>=2.9.1
 torchvision>=0.21.0
 transformers>=4.48.1
 opencv-python>=4.2.0.32

{autoforge-1.9.2.2 → autoforge-1.9.3}/src/autoforge/Helper/OutputHelper.py

@@ -906,19 +906,52 @@ def _save_stl_with_manifold_fix(triangles, filename):
     buffer.write(struct.pack("<I", num_triangles))
     buffer.write(stl_data.tobytes())
     buffer.seek(0)
-    
-    # Load with trimesh and fix manifold issues
-    try:
-        mesh = trimesh.load(buffer, file_type="stl")
-        mesh.merge_vertices()
-        mesh.remove_degenerate_faces()
-        mesh.remove_duplicate_faces()
-        mesh.fill_holes()
-        mesh.fix_normals()
-        mesh.export(filename)
-    except Exception as e:
-        print(f"Warning: Error fixing manifold for {filename}: {e}")
-        # Fall back to saving without manifold fix
-        with open(filename, "wb") as f:
-            buffer.seek(0)
-            f.write(buffer.read())
+
+    with open(filename, "wb") as f:
+        buffer.seek(0)
+        f.write(buffer.read())
+    
+    # # Load with trimesh and fix manifold issues
+    # try:
+    #     mesh = trimesh.load(buffer, file_type="stl")
+    #     mesh.merge_vertices()
+    #
+    #     # Try available degenerate-face removal utilities across trimesh versions
+    #     try:
+    #         if hasattr(mesh, "remove_degenerate_faces"):
+    #             mesh.remove_degenerate_faces()
+    #         elif hasattr(trimesh.repair, "remove_degenerate_faces"):
+    #             trimesh.repair.remove_degenerate_faces(mesh)
+    #         elif hasattr(trimesh.repair, "remove_degenerate_triangles"):
+    #             trimesh.repair.remove_degenerate_triangles(mesh)
+    #     except Exception as repair_err:
+    #         print(f"Warning: Skipped degenerate-face removal for {filename}: {repair_err}")
+    #
+    #     # Remove duplicate faces if the method exists
+    #     try:
+    #         if hasattr(mesh, "remove_duplicate_faces"):
+    #             mesh.remove_duplicate_faces()
+    #         elif hasattr(trimesh.repair, "remove_duplicate_faces"):
+    #             trimesh.repair.remove_duplicate_faces(mesh)
+    #     except Exception as repair_err:
+    #         print(f"Warning: Skipped duplicate-face removal for {filename}: {repair_err}")
+    #
+    #     # Fill holes and fix normals when available
+    #     try:
+    #         if hasattr(mesh, "fill_holes"):
+    #             mesh.fill_holes()
+    #     except Exception as repair_err:
+    #         print(f"Warning: Skipped hole filling for {filename}: {repair_err}")
+    #
+    #     try:
+    #         mesh.fix_normals()
+    #     except Exception as repair_err:
+    #         print(f"Warning: Skipped normal fixing for {filename}: {repair_err}")
+    #
+    #     mesh.export(filename)
+    # except Exception as e:
+    #     print(f"Warning: Error fixing manifold for {filename}: {e}")
+    #     # Fall back to saving without manifold fix
+    #     with open(filename, "wb") as f:
+    #         buffer.seek(0)
+    #         f.write(buffer.read())

{autoforge-1.9.2.2 → autoforge-1.9.3}/src/autoforge/Helper/PruningHelper.py

@@ -5,6 +5,7 @@ import torch
 from tqdm import tqdm
 from joblib import Parallel, delayed
 import threading
+import numpy as np
 
 from autoforge.Helper.OptimizerHelper import composite_image_disc
 from autoforge.Loss.LossFunctions import compute_loss
@@ -864,3 +865,101 @@ def optimise_swap_positions(
 
     outer_tbar.close()
     return best_dg
+
+
+def _compute_loss_for_heightmap(
+    optimizer: FilamentOptimizer,
+    disc_global: torch.Tensor,
+    disc_height: torch.Tensor,
+    *,
+    custom_height_logits: torch.Tensor | None = None,
+) -> float:
+    """Compute discrete loss for a given discrete height map/global assignment.
+
+    If ``custom_height_logits`` is provided, they are used directly (bypassing
+    discretization) when compositing the image; otherwise the optimizer's current
+    best height logits are used.
+    """
+    logits_for_disc = disc_to_logits(disc_global, optimizer.material_colors.shape[0], big_pos=1e5)
+    with _gpu_lock, torch.no_grad():
+        out_im = optimizer.get_best_discretized_image(
+            custom_height_logits=custom_height_logits,
+            custom_global_logits=logits_for_disc,
+        )
+        return compute_loss(comp=out_im, target=optimizer.target).item()
+
+
+def _median_without_outliers(window: np.ndarray, threshold: float, max_outliers: int = 2) -> tuple[float, bool, int]:
+    """Return replacement for center pixel if it is a spike; also report spike count.
+
+    A spike is when the center pixel is >= ``threshold`` above the window median and
+    the window has at most ``max_outliers`` such outliers. Returns (replacement_value,
+    is_spike, spike_count).
+    """
+    med = np.median(window)
+    outlier_mask = window - med >= threshold
+    spike_count = int(outlier_mask.sum())
+    center_is_outlier = bool(outlier_mask[1, 1])
+    if not center_is_outlier or spike_count == 0 or spike_count > max_outliers:
+        return float(window[1, 1]), False, 0
+    non_outliers = window[~outlier_mask]
+    if non_outliers.size == 0:
+        return float(window[1, 1]), False, 0
+    replacement = float(np.median(non_outliers))
+    return replacement, True, spike_count
+
+
+def remove_height_spikes(
+    disc_height: torch.Tensor, threshold_layers: int = 3, max_outliers: int = 2
+) -> tuple[torch.Tensor, int]:
+    """Remove tall spikes in a discrete height map using a 3x3 neighborhood.
+
+    A spike occurs only when the *center* pixel of a 3x3 window is at least
+    ``threshold_layers`` higher than the window median and the window has at most
+    two such outliers. Only those center pixels are replaced with the median of
+    the non-outlier values. Operates in layer units (0..max_layers).
+
+    Returns:
+        cleaned_height: torch.Tensor matching ``disc_height`` shape/device.
+        spike_count: int number of pixels corrected.
+    """
+    if disc_height.ndim != 2:
+        raise ValueError("disc_height must be 2D [H,W]")
+    if threshold_layers <= 0:
+        return disc_height, 0
+    if max_outliers <= 0:
+        return disc_height, 0
+
+    dh_np = disc_height.detach().cpu().numpy()
+    H, W = dh_np.shape
+    if H < 3 or W < 3:
+        return disc_height, 0
+
+    out = dh_np.copy()
+    spikes = 0
+    tbar = tqdm(range(1, H - 1), desc="Removing height spikes", leave=False)
+    for y in tbar:
+        for x in range(1, W - 1):
+            window = dh_np[y - 1 : y + 2, x - 1 : x + 2]
+            med_all = np.median(window)
+            mask = window - med_all >= threshold_layers
+            spike_count = int(mask.sum())
+            if not mask[1, 1] or spike_count == 0 or spike_count > max_outliers:
+                continue
+
+            non_outliers = window[~mask]
+            if non_outliers.size == 0:
+                continue
+            med_non_outliers = float(np.median(non_outliers))
+            center_val = float(window[1, 1])
+            if center_val < med_non_outliers + threshold_layers:
+                continue
+
+            out[y, x] = med_non_outliers  # only replace the center spike
+            spikes += 1
+
+    if spikes == 0:
+        return disc_height, 0
+
+    cleaned = torch.tensor(out, dtype=disc_height.dtype, device=disc_height.device)
+    return cleaned, spikes