File size: 13,997 Bytes
9affda1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
import os
import sys
import random
import torch
import numpy as np
import torch.nn.functional as F
from argparse import ArgumentParser
from core.registry import register_method
from core.base_method import BaseMethod

sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '../../improvingadc_official')))
from utils.loss_utils import l1_loss, ssim
from gaussian_renderer import render as native_render
from scene import Scene, GaussianModel
from arguments import ModelParams, PipelineParams, OptimizationParams

@register_method("improvingadc")
class ImprovingADCWrapper(BaseMethod):
    def __init__(self, dataset_config, hyperparams):
        self.parser = ArgumentParser()
        self.lp = ModelParams(self.parser)
        self.op = OptimizationParams(self.parser)
        self.pp = PipelineParams(self.parser)
        self.args = self.parser.parse_args([])

        self.args.source_path = dataset_config["source_path"]
        self.args.model_path = dataset_config["model_path"]
        self.args.eval = True
        self.args.resolution = dataset_config.get("resolution", 1)
        self.track_decoupling = hyperparams.get("track_decoupling", False)

        self.dataset = self.lp.extract(self.args)
        self.opt = self.op.extract(self.args)
        self.pipe = self.pp.extract(self.args)

        self.gaussians = GaussianModel(self.dataset.sh_degree)
        self.gaussians.set_dl(folder=self.args.model_path, log_frq=50, param_log=False)
        self.scene = Scene(self.dataset, self.gaussians)
        self.gaussians.training_setup(self.opt)

        bg_color = [1, 1, 1] if self.dataset.white_background else [0, 0, 0]
        self.background = torch.tensor(bg_color, dtype=torch.float32, device="cuda")
        self.viewpoint_stack = self.scene.getTrainCameras().copy()
        
        self.last_n_gaussians = len(self.gaussians.get_xyz)

        self.dense_iters = ((self.opt.densify_until_iter - self.opt.densify_from_iter) // self.opt.densification_interval)
        self.start_thresh = 0.0001
        self.end_thresh = 0.0004
        self.dense_factor = pow((self.end_thresh / self.start_thresh), pow(self.dense_iters, -1))
        self.current_dense_thresh = self.start_thresh
        self.per_gaussian_alpha = None

    def train_iteration(self, step):
        self.gaussians.update_learning_rate(step)
        self.gaussians.dl.iteration = step
        self.gaussians.dl.log_in_file(self.gaussians)
        if step % 1000 == 0:
            self.gaussians.oneupSHdegree()
            
        if not self.viewpoint_stack:
            self.viewpoint_stack = self.scene.getTrainCameras().copy()
            
        viewpoint_cam = self.viewpoint_stack.pop(random.randint(0, len(self.viewpoint_stack) - 1))

        bg = torch.rand((3), device="cuda") if self.opt.random_background else self.background
        
        render_pkg = native_render(viewpoint_cam, self.gaussians, self.pipe, bg, depth_threshold=self.opt.depth_threshold * self.scene.old_extent)
        image = render_pkg["render"]
        viewspace_point_tensor = render_pkg["viewspace_points"]
        visibility_filter = render_pkg["visibility_filter"]
        radii = render_pkg["radii"]
        
        gt_image = viewpoint_cam.original_image.cuda()

        fake_color = torch.zeros_like(self.gaussians._xyz, requires_grad=True)
        fake_render = native_render(viewpoint_cam, self.gaussians, self.pipe, bg, override_color=fake_color, depth_threshold=self.opt.depth_threshold * self.scene.old_extent)["render"]
        fake_loss = torch.sum(fake_render.view(-1))
        
        Ll1 = l1_loss(image, gt_image)
        ssim_value = ssim(image, gt_image)
        loss_target = (1.0 - self.opt.lambda_dssim) * Ll1
        loss_parasitic = self.opt.lambda_dssim * (1.0 - ssim_value)
        
        loss = loss_target + loss_parasitic + fake_loss

        grad_cos_sim = 0.0
        parasitic_ratio = 0.0
        stats = {}

        if self.track_decoupling and step % 100 == 0:
            self.gaussians.optimizer.zero_grad(set_to_none=True)
            loss_target.backward(retain_graph=True)
            grad_target_xyz = self.gaussians._xyz.grad.clone() if self.gaussians._xyz.grad is not None else torch.zeros_like(self.gaussians._xyz)
            
            param_groups_map = {
                "spatial": [self.gaussians._xyz],
                "geometry": [self.gaussians._scaling, self.gaussians._rotation],
                "opacity": [self.gaussians._opacity],
                "appearance": [self.gaussians._features_dc, self.gaussians._features_rest],
            }

            grads_target = {}
            for group_name, params in param_groups_map.items():
                group_grads = []
                for p in params:
                    if p.grad is not None:
                        state = self.gaussians.optimizer.state.get(p, {})
                        v = state.get("exp_avg_sq", torch.zeros_like(p.grad))
                        for pg in self.gaussians.optimizer.param_groups:
                            if pg['params'][0] is p:
                                lr = pg['lr']
                                break
                        else:
                            lr = 1e-4
                        u = (lr / (torch.sqrt(v) + 1e-8)) * p.grad.clone()
                        group_grads.append(u.reshape(-1))
                if group_grads:
                    grads_target[group_name] = torch.cat(group_grads)

            self.gaussians.optimizer.zero_grad(set_to_none=True)
            loss_parasitic.backward(retain_graph=True)
            grad_parasitic_xyz = self.gaussians._xyz.grad.clone() if self.gaussians._xyz.grad is not None else torch.zeros_like(self.gaussians._xyz)
            
            grads_parasitic = {}
            for group_name, params in param_groups_map.items():
                group_grads = []
                for p in params:
                    if p.grad is not None:
                        state = self.gaussians.optimizer.state.get(p, {})
                        v = state.get("exp_avg_sq", torch.zeros_like(p.grad))
                        for pg in self.gaussians.optimizer.param_groups:
                            if pg['params'][0] is p:
                                lr = pg['lr']
                                break
                        else:
                            lr = 1e-4
                        u = (lr / (torch.sqrt(v) + 1e-8)) * p.grad.clone()
                        group_grads.append(u.reshape(-1))
                if group_grads:
                    grads_parasitic[group_name] = torch.cat(group_grads)

            valid_mask = (torch.norm(grad_target_xyz, dim=1) > 0) & (torch.norm(grad_parasitic_xyz, dim=1) > 0)
            if valid_mask.any():
                grad_cos_sim = float(F.cosine_similarity(grad_target_xyz[valid_mask], grad_parasitic_xyz[valid_mask], dim=1).mean())
            parasitic_ratio = float(torch.norm(grad_parasitic_xyz, dim=1).mean() / (torch.norm(grad_target_xyz, dim=1).mean() + 1e-7))

            for group_name in param_groups_map:
                gt = grads_target.get(group_name)
                gp = grads_parasitic.get(group_name)
                if gt is not None and gp is not None and gt.norm() > 0 and gp.norm() > 0:
                    cos = float(F.cosine_similarity(gt.unsqueeze(0), gp.unsqueeze(0)))
                    r = float(gp.norm() / (gt.norm() + gp.norm() + 1e-7))
                    ti = r * max(0.0, -cos)
                else:
                    ti = 0.0
                stats[f"sti_{group_name}"] = ti

            self.gaussians.optimizer.zero_grad(set_to_none=True)
            loss.backward()
        else:
            loss.backward()

        if self.per_gaussian_alpha is not None:
            self.per_gaussian_alpha += torch.mean(fake_color.grad, dim=1)
        else:
            self.per_gaussian_alpha = torch.mean(fake_color.grad, dim=1)

        with torch.no_grad():
            if step < self.opt.densify_until_iter:
                self.gaussians.max_radii2D[visibility_filter] = torch.max(self.gaussians.max_radii2D[visibility_filter], radii[visibility_filter])
                self.gaussians.add_densification_stats(viewspace_point_tensor, visibility_filter, render_pkg["pixels"])
                
                if step > self.opt.densify_from_iter and step % self.opt.densification_interval == 0:
                    size_threshold = 20 if step > self.opt.opacity_reset_interval else None
                    self.current_dense_thresh *= self.dense_factor
                    self.gaussians.densify_and_prune(self.current_dense_thresh, 0.005, self.scene.cameras_extent, size_threshold, metric=self.per_gaussian_alpha)
                    self.per_gaussian_alpha = None
                
                if step % self.opt.opacity_reset_interval == 0 or (self.dataset.white_background and step == self.opt.densify_from_iter):
                    self.gaussians.reset_opacity()

            if step < self.opt.iterations:
                self.gaussians.optimizer.step()
                self.gaussians.optimizer.zero_grad(set_to_none=True)

        num_gaussians = self.gaussians.get_xyz.shape[0]
        alpha_mean_val = float(self.per_gaussian_alpha.mean()) if self.per_gaussian_alpha is not None else 0.0

        metrics = {
            "loss": float(loss), "loss_l1": float(loss_target), "loss_ssim": float(loss_parasitic),
            "num_gaussians": int(num_gaussians), "delta_N": int(num_gaussians - self.last_n_gaussians),
            "peak_vram_GB": float(torch.cuda.max_memory_allocated() / (1024 ** 3)),
            "grad_cos_sim": float(grad_cos_sim), "parasitic_ratio": float(parasitic_ratio),
            "dense_threshold_current": float(self.current_dense_thresh),
            "per_gaussian_alpha_mean": alpha_mean_val
        }
        metrics.update(stats)
        self.last_n_gaussians = num_gaussians
        
        histograms = {}
        if step % 1000 == 0:
            histograms["opacity"] = torch.sigmoid(self.gaussians._opacity).clone().detach()
            scales = torch.exp(self.gaussians._scaling).clone().detach()
            histograms["scaling"] = scales
            scales_2d = scales[:, :2] if scales.shape[1] >= 2 else scales
            gamma = scales_2d.max(dim=-1)[0] / (scales_2d.min(dim=-1)[0] + 1e-7)
            histograms["anisotropy"] = gamma
            histograms["sh_dc_mag"] = self.gaussians._features_dc.detach().norm(dim=-1)

        return metrics, histograms
        
    def render(self, camera):
        with torch.no_grad():
            bg = torch.tensor([1, 1, 1] if self.dataset.white_background else [0, 0, 0], dtype=torch.float32, device="cuda")
            render_pkg = native_render(camera, self.gaussians, self.pipe, bg)
        return {"image": render_pkg["render"], "depth": render_pkg.get("depth", None)}

    def save(self, save_dir, step):
        self.scene.save(step)

    def load(self, model_path, iteration):
        self.gaussians.load_ply(os.path.join(model_path, 'point_cloud', f'iteration_{iteration}', 'point_cloud.ply'))

    def get_spatial_centers(self):
        return self.gaussians._xyz

    def compute_physical_metrics(self, cameras=None):
        metrics = {}
        with torch.no_grad():
            raw_scales = self.gaussians._scaling
            scales = torch.exp(raw_scales)
            
            scales_2d = scales[:, :2] if scales.dim() > 1 and scales.shape[1] >= 2 else scales.unsqueeze(-1).expand(-1, 2)
            
            max_S, _ = torch.max(scales_2d, dim=1)
            min_S, _ = torch.min(scales_2d, dim=1)
            gamma = max_S / (min_S + 1e-7)
            
            metrics["gamma_median"] = float(torch.median(gamma))
            metrics["gamma_90th_percentile"] = float(torch.quantile(gamma, 0.90))
            metrics["scale_mean"] = float(torch.mean(scales_2d))
            metrics["alpha_mean"] = float(torch.mean(torch.sigmoid(self.gaussians._opacity)))
            
            dc, rest = self.gaussians._features_dc, self.gaussians._features_rest
            if rest is not None and rest.shape[1] > 0:
                metrics["sh_energy_ratio"] = float(rest.norm(dim=-1).mean() / (dc.norm(dim=-1).mean() + 1e-7))
                
            if cameras is not None and len(cameras) > 0:
                view_dirs = []
                for c in cameras:
                    view_dirs.append(c.world_view_transform[:3, 2].tolist())
                view_dirs = F.normalize(torch.tensor(view_dirs, dtype=torch.float32, device="cuda"), dim=1)

                rots = F.normalize(self.gaussians._rotation.clone(), dim=1)
                w, x, y, z = rots.unbind(dim=-1)
                normals = F.normalize(torch.stack([2*(x*z + w*y), 2*(y*z - w*x), 1-2*(x*x + y*y)], dim=-1), dim=1)
                
                max_cos, _ = torch.max(torch.abs(torch.matmul(normals, view_dirs.T)), dim=1)
                metrics["billboard_bias_ratio"] = float((max_cos > 0.90).float().mean())

        return metrics

    def evaluate_spatial_field(self, query_points: torch.Tensor, cameras=None) -> torch.Tensor:
        with torch.no_grad():
            V = query_points.shape[0]
            densities = torch.zeros(V, device="cuda")
            xyz, opacities = self.gaussians._xyz, torch.sigmoid(self.gaussians._opacity).squeeze()
            scales = torch.exp(self.gaussians._scaling)
            sigma_sq = (scales[:, :2].max(dim=1)[0].pow(2)) if scales.shape[1] >= 2 else scales.squeeze().pow(2)
            
            N_gaussians = xyz.shape[0]
            chunk_size = max(1, 30_000_000 // (N_gaussians + 1))
            for i in range(0, V, chunk_size):
                end = min(i + chunk_size, V)
                dist_sq = torch.cdist(query_points[i:end], xyz, p=2).pow(2)
                weights = torch.exp(-0.5 * dist_sq / (sigma_sq.unsqueeze(0) + 1e-7))
                densities[i:end] = torch.sum(weights * opacities.unsqueeze(0), dim=1)
            return densities