Ewald
a4154ccf05
- Fix batch bypass tuple handling, memory budget device-awareness, grain RNG device, roto payload limits, preview temp-file cleanup, and route double-registration. - Add native VIDEO output (fps/audio) to effect nodes that previously only returned IMAGE/MASK when fed a VIDEO input. - Split the oversized src/preview/ops/geometry.ts and add a starting golden-fixture test harness (spherize, transform) comparing backend and frontend math. - Remove ImageOpsAppendV2 (Autogrow) and ImageOpsLensArtifacts; inline Append's helper back into its single caller. - Collapse the Color Correct curve widget behind a toggle so the node no longer grows too tall. - Fix Crop/generic-node video preview playback running at the RAF rate instead of the source video's real fps. - Fix Crop losing its saved output size after a workflow reload: syncCropWidgets recomputed height from width and a preset aspect ratio on every resync (including onConfigure on load), discarding the restored value. It now only refreshes aspect-ratio bookkeeping on a resync.
93 lines
3.2 KiB
Python
93 lines
3.2 KiB
Python
"""Unit tests for ImageOpsSpherize verifying mask grid warping and bypass."""
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
from pathlib import Path
|
|
|
|
import pytest
|
|
import torch
|
|
|
|
from nodes.spherize import ImageOpsSpherize, _spherize_frame
|
|
|
|
|
|
def _unwrap(result):
|
|
if isinstance(result, dict):
|
|
return result["result"]
|
|
return result
|
|
|
|
|
|
def test_spherize_frame_matches_golden_fixture_used_for_frontend_parity():
|
|
"""Regression guard for tests/golden/spherize.json, which
|
|
tests/frontend/spherize.test.cjs uses to check applySpherize for parity.
|
|
If this test needs updating, regenerate the golden and re-check the JS side."""
|
|
golden_path = Path(__file__).resolve().parents[1] / "golden" / "spherize.json"
|
|
golden = json.loads(golden_path.read_text(encoding="utf-8"))
|
|
meta = golden["_meta"]
|
|
width, height = meta["width"], meta["height"]
|
|
|
|
frame = torch.tensor(golden["input_rgb"], dtype=torch.float32).unsqueeze(0) / 255.0
|
|
out = _spherize_frame(frame, **golden["params"])
|
|
actual = (out[0] * 255.0).round().clamp(0, 255).byte()
|
|
expected = torch.tensor(golden["expected_rgb"], dtype=torch.uint8)
|
|
|
|
assert actual.shape == (height, width, 3)
|
|
max_diff = (actual.int() - expected.int()).abs().max().item()
|
|
# tolerance=1: the fixture round-trips through uint8 storage, so re-deriving
|
|
# the input tensor from it reintroduces ~1/255 of quantization noise versus
|
|
# the exact floats used when the golden was generated.
|
|
assert max_diff <= 1, f"_spherize_frame no longer matches its own golden fixture (max diff {max_diff})"
|
|
|
|
|
|
class TestImageOpsSpherize:
|
|
def test_spherize_bypass_preserves_resolution(self):
|
|
node = ImageOpsSpherize()
|
|
# Input image is 120x80
|
|
image = torch.rand(1, 80, 120, 3, dtype=torch.float32)
|
|
mask = torch.rand(1, 80, 120, dtype=torch.float32)
|
|
|
|
# Custom size is set to 256x256, but bypass=True should NEVER resize
|
|
result = _unwrap(node.execute(
|
|
image=image,
|
|
mask=mask,
|
|
bypass=True,
|
|
size_mode="custom",
|
|
width=256,
|
|
height=256,
|
|
))
|
|
|
|
out_image = result[0]
|
|
out_mask = result[1]
|
|
|
|
assert out_image.shape == (1, 80, 120, 3)
|
|
assert out_mask.shape == (1, 80, 120)
|
|
assert torch.allclose(out_image, image)
|
|
assert torch.allclose(out_mask, mask)
|
|
|
|
def test_spherize_mask_warped_with_image(self):
|
|
node = ImageOpsSpherize()
|
|
# Put a spot at center
|
|
image = torch.zeros(1, 64, 64, 3, dtype=torch.float32)
|
|
image[0, 30:34, 30:34, :] = 1.0
|
|
|
|
mask = torch.zeros(1, 64, 64, dtype=torch.float32)
|
|
mask[0, 30:34, 30:34] = 1.0
|
|
|
|
result = _unwrap(node.execute(
|
|
image=image,
|
|
mask=mask,
|
|
mode="spherize",
|
|
strength=1.0,
|
|
size_mode="from_input",
|
|
))
|
|
|
|
out_image = result[0]
|
|
out_mask = result[1]
|
|
|
|
assert out_image.shape == (1, 64, 64, 3)
|
|
assert out_mask.shape == (1, 64, 64)
|
|
assert torch.isfinite(out_image).all()
|
|
assert torch.isfinite(out_mask).all()
|
|
# Center spot should still be present and warped in both image and mask
|
|
assert out_image[0, 32, 32, 0] > 0.5
|
|
assert out_mask[0, 32, 32] > 0.5
|