SVG在网页浏览器中渲染正常，但pydiffvg中因缺失stroke属性等问题渲染异常的解决方法问询

我最近在用pydiffvg渲染SVG时遇到了个头疼的问题——当SVG里缺少stroke-width这类描边属性时，要么渲染失败，要么颜色不对、元素缺失，但同样的SVG在浏览器或Inkscape里却能完美显示，看起来它们会自动补全这些缺失的属性。

我已经尝试过这些方法，但都没解决问题：

• 给解析器设置默认的描边值（黑色或透明）
• 当stroke-width未定义时，强制设为1.0

但即使这么做了，pydiffvg还是表现不佳，而浏览器却能轻松处理。我想知道：浏览器处理不完整SVG时到底做了什么不同的操作？我该怎么复现这种行为，让pydiffvg能正确渲染这个SVG？

我使用的SVG代码 #

<?xml version="1.0" encoding="utf-8"?>
<!-- Uploaded to: SVG Repo, www.svgrepo.com, Generator: SVG Repo Mixer Tools -->
<svg width="800px" height="800px" viewBox="0 0 1024 1024" class="icon"  version="1.1" xmlns="http://www.w3.org/2000/svg"><path d="M301.226643 1002.666249A34.773319 34.773319 0 0 1 266.666658 968.106263V399.999833c0-45.226648 71.253304-152.53327 116.266618-209.706579l4.479998-5.759998V109.439954h248.959896v75.093302l4.479999 5.759998C686.079816 247.466564 757.33312 354.773186 757.33312 399.999833v568.10643A34.773319 34.773319 0 0 1 722.559801 1002.666249z" fill="#F05071" /><path d="M615.039846 130.773279V191.99992l9.173329 11.519995c53.759978 68.053305 111.78662 162.559932 111.786621 196.479918v568.10643a13.439994 13.439994 0 0 1-13.226662 13.226661H301.226643a13.439994 13.439994 0 0 1-13.226661-13.226661V399.999833c0-33.706653 58.026642-127.999947 111.78662-196.479918L408.959932 191.99992V130.773279h206.293247m42.666649-42.666649H366.293283V177.066593S245.333333 330.453196 245.333333 399.999833v568.10643A55.89331 55.89331 0 0 0 301.226643 1023.999573h421.333158A55.89331 55.89331 0 0 0 778.666444 968.106263V399.999833c0-69.546638-120.95995-222.93324-120.959949-222.93324V88.10663z" fill="#5C2D51" /><path d="M321.706635 21.333324l380.373175 0 0 106.666623-380.373175 0 0-106.666623Z" fill="#9FDBAD" /><path d="M680.959818 42.666649v63.999973H343.039959V42.666649h337.919859m15.146661-42.666649H327.893299a27.519989 27.519989 0 0 0-27.519989 27.519989v94.293294A27.519989 27.519989 0 0 0 327.893299 149.333271h367.999847a27.519989 27.519989 0 0 0 27.519988-27.519988V27.519989A27.519989 27.519989 0 0 0 695.893146 0z" fill="#5C2D51" /><path d="M266.666658 469.333138h490.666462v341.333191H266.666658z" fill="#9FDBAD" /><path d="M735.999796 490.666462v298.666542H287.999982V490.666462h447.999814m42.666648-42.666649H245.333333v383.99984h533.333111V447.999813z" fill="#5C2D51" /></svg>

渲染效果对比 #

这是SVG应该呈现的正确效果

这是pydiffvg渲染后的结果（忽略方向，我已经手动调整过了）

我的渲染代码 #

if shape_type == "custom" and svg_path:
                try:
                    _, _, parsed_shapes, parsed_shape_groups = svg_to_scene(svg_path)

                    # Filter out invalid shapes
                    parsed_shapes = [
                        shape for shape in parsed_shapes
                        if hasattr(shape, "points") or hasattr(shape, "center")
                    ]

                    print(f"Parsed {len(parsed_shapes)} valid shapes from {svg_path}.")


                    if not parsed_shapes:
                        print(f"Warning: No valid shapes were parsed from {svg_path}.")
                        continue

                    scale_factor = shape_info.get("size", min(self.canvas_width, self.canvas_height) * 0.01)

                    # Process each logical group based on count
                    for g in range(count):
                        group_shapes = []
                        for i, (shape, shape_group) in enumerate(zip(parsed_shapes, parsed_shape_groups)):
                            if hasattr(shape, "points"):
                                shape.points = shape.points.to(self.device) * scale_factor
                                shape.initial_points = shape.points.clone()
                            elif hasattr(shape, "center"):
                                shape.center = shape.center.to(self.device) * scale_factor
                                shape.initial_points = shape.center.unsqueeze(0).clone()

                            # Set default colors if missing
                            shape_group.fill_color = (
                                shape_group.fill_color.clone().to(self.device)
                                if shape_group.fill_color is not None
                                else torch.tensor([1.0, 1.0, 1.0, 1.0], device=self.device)
                            )
                            shape_group.stroke_color = (
                                shape_group.stroke_color.clone().to(self.device)
                                if shape_group.stroke_color is not None
                                else shape_group.fill_color.clone().to(self.device) # Default to fill color
                            )

                            if isinstance(shape, pydiffvg.Path):
                                new_shape = self.copy_path(shape)
                            elif isinstance(shape, pydiffvg.Polygon):
                                new_shape = pydiffvg.Polygon(
                                    points=shape.points.clone(),
                                    is_closed=shape.is_closed,
                                )
                            elif isinstance(shape, pydiffvg.Circle):
                                new_shape = pydiffvg.Circle(
                                    radius=shape.radius.clone(),
                                    center=shape.center.clone(),
                                )
                            else:
                                print(f"Warning: Unsupported shape type {type(shape)}.")
                                continue

                            new_shape.initial_points = shape.initial_points.clone()
                            group_shapes.append(new_shape)
                            self.shapes.append(new_shape)

                            new_shape_group = pydiffvg.ShapeGroup(
                                shape_ids=torch.tensor([len(self.shapes) - 1], dtype=torch.int32).to(self.device),
                                fill_color=shape_group.fill_color,
                                stroke_color=shape_group.stroke_color,
                            )
                            self.shape_groups.append(new_shape_group)

                        self.shapes_by_logical_group.append(group_shapes)

                        # Initialize position and orientation
                        position = torch.tensor(
                            [
                                center_x + (torch.rand(1).item() - 0.5) * 2 * max_radius,
                                center_y + (torch.rand(1).item() - 0.5) * 2 * max_radius,
                            ],
                            requires_grad=True,
                            device=self.device,
                        )
                        orientation = torch.tensor(
                            [torch.rand(1).item() * 360],
                            requires_grad=True,
                            device=self.device,
                        )

                        # Apply transformations to group shapes
                        for shape in group_shapes:
                            if hasattr(shape, "initial_points"):
                                angle = torch.deg2rad(orientation[0])
                                rotation_matrix = torch.tensor([
                                    [torch.cos(angle), -torch.sin(angle)],
                                    [torch.sin(angle), torch.cos(angle)]
                                ]).to(self.device)

                                transformed_points = torch.matmul(shape.initial_points, rotation_matrix.T) + position
                                shape.points = transformed_points.clone()
                            elif isinstance(shape, pydiffvg.Circle):
                                shape.center = position.clone()

                        self.parameters.position.append(position)
                        self.parameters.orientation.append(orientation)

                        logical_group_count += 1

                except Exception as e:
                    print(f"Error processing custom SVG {svg_path}: {e}")
                continue

有没有大佬能指点下，该怎么预处理或解析这类SVG，让pydiffvg能正确渲染它？感激不尽！

备注：内容来源于stack exchange，提问作者Ravish Jha