gsvvcompressor 1.2.0__cp310-cp310-win_amd64.whl

gsvvcompressor/xyz/size.py

@@ -0,0 +1,44 @@
+"""
+Quantization step size calculation.
+"""
+
+from typing import Optional
+
+
+def compute_step_size(
+    dense_scale: float,
+    alpha: float = 0.2,
+    min_step: Optional[float] = None,
+    max_step: Optional[float] = None,
+) -> float:
+    """
+    Compute quantization step size from dense scale.
+
+    The step size is computed as:
+        step_size = alpha * dense_scale
+
+    Args:
+        dense_scale: The dense region scale estimate (d_dense), typically from
+            compute_dense_scale().
+        alpha: Scaling factor for step size (typical range: 0.1 to 0.25).
+            Smaller values preserve more detail but increase data size.
+        min_step: Optional minimum step size to prevent extremely fine quantization.
+        max_step: Optional maximum step size to prevent overly coarse quantization.
+
+    Returns:
+        The quantization step size (delta).
+
+    Example:
+        >>> from cevv.xyz import compute_nn_distances, compute_dense_scale, compute_step_size
+        >>> nn_distances = compute_nn_distances(points, k=1, sample_size=10000)
+        >>> dense_scale = compute_dense_scale(nn_distances, quantile=0.05)
+        >>> step_size = compute_step_size(dense_scale, alpha=0.2)
+    """
+    step_size = alpha * dense_scale
+
+    if min_step is not None:
+        step_size = max(step_size, min_step)
+    if max_step is not None:
+        step_size = min(step_size, max_step)
+
+    return step_size

gsvvcompressor/xyz/twopass.py

@@ -0,0 +1 @@
+# TODO: two pass xyz interframe encoder/decoder, produce codebook on the whole 3DGS seq in pass one