| import argparse, json |
| from pathlib import Path |
| import torch |
| from arguments import ModelParams, PipelineParams, OptimizationParams |
| from scene import Scene, FourierModel |
|
|
| parser = argparse.ArgumentParser() |
| lp = ModelParams(parser); op = OptimizationParams(parser); pp = PipelineParams(parser) |
| parser.add_argument('--load_iteration', type=int, default=6000) |
| args = parser.parse_args() |
| dataset = lp.extract(args) |
| gaussians = FourierModel(dataset.sh_degree, fourier_K=dataset.fourier_K, |
| use_beta_color=dataset.beta_color, sb_lobes=dataset.sb_lobes) |
| scene = Scene(dataset, gaussians, load_iteration=args.load_iteration, shuffle=False) |
| out = Path(dataset.model_path) / 'lod_models' |
| out.mkdir(parents=True, exist_ok=True) |
| full_r = gaussians._radii_logits.detach().clone() |
| full_p = gaussians._phases_raw.detach().clone() |
| full_k = gaussians.fourier_K |
| rows = [] |
| for k in range(1, full_k+1): |
| gaussians._radii_logits = torch.nn.Parameter(full_r[:, :k].clone().requires_grad_(False)) |
| gaussians._phases_raw = torch.nn.Parameter(full_p[:, :k].clone().requires_grad_(False)) |
| gaussians.fourier_K = k |
| gaussians.active_K = k |
| path = out / f'fourier_k{k}.ply' |
| gaussians.save_ply(str(path)) |
| rows.append({'K':k,'path':str(path),'bytes':path.stat().st_size,'primitives':int(gaussians.get_xyz.shape[0])}) |
| gaussians._radii_logits = torch.nn.Parameter(full_r) |
| gaussians._phases_raw = torch.nn.Parameter(full_p) |
| gaussians.fourier_K = full_k |
| gaussians.active_K = full_k |
| (out/'sizes.json').write_text(json.dumps(rows,indent=2)) |
| print(json.dumps(rows,indent=2)) |
|
|
