train_mlp_dnerf.py

"""
Copyright (c) 2022 Ruilong Li, UC Berkeley.
"""

import argparse
import math
import pathlib
import time

import imageio
import numpy as np
import torch
import torch.nn.functional as F
import tqdm
from datasets.dnerf_synthetic import SubjectLoader
from radiance_fields.mlp import DNeRFRadianceField
from utils import render_image, set_random_seed

from nerfacc import ContractionType, OccupancyGrid

if __name__ == "__main__":

    device = "cuda:0"
    set_random_seed(42)

    parser = argparse.ArgumentParser()
    parser.add_argument(
        "--data_root",
        type=str,
        default=str(pathlib.Path.cwd() / "data/dnerf"),
        help="the root dir of the dataset",
    )
    parser.add_argument(
        "--train_split",
        type=str,
        default="train",
        choices=["train"],
        help="which train split to use",
    )
    parser.add_argument(
        "--scene",
        type=str,
        default="lego",
        choices=[
            # dnerf
            "bouncingballs",
            "hellwarrior",
            "hook",
            "jumpingjacks",
            "lego",
            "mutant",
            "standup",
            "trex",
        ],
        help="which scene to use",
    )
    parser.add_argument(
        "--aabb",
        type=lambda s: [float(item) for item in s.split(",")],
        default="-1.5,-1.5,-1.5,1.5,1.5,1.5",
        help="delimited list input",
    )
    parser.add_argument(
        "--test_chunk_size",
        type=int,
        default=8192,
    )
    parser.add_argument("--cone_angle", type=float, default=0.0)
    args = parser.parse_args()

    render_n_samples = 1024

    # setup the scene bounding box.
    contraction_type = ContractionType.AABB
    scene_aabb = torch.tensor(args.aabb, dtype=torch.float32, device=device)
    near_plane = None
    far_plane = None
    render_step_size = (
        (scene_aabb[3:] - scene_aabb[:3]).max()
        * math.sqrt(3)
        / render_n_samples
    ).item()

    # setup the radiance field we want to train.
    max_steps = 30000
    grad_scaler = torch.cuda.amp.GradScaler(1)
    radiance_field = DNeRFRadianceField().to(device)
    optimizer = torch.optim.Adam(radiance_field.parameters(), lr=5e-4)
    scheduler = torch.optim.lr_scheduler.MultiStepLR(
        optimizer,
        milestones=[
            max_steps // 2,
            max_steps * 3 // 4,
            max_steps * 5 // 6,
            max_steps * 9 // 10,
        ],
        gamma=0.33,
    )
    # setup the dataset
    target_sample_batch_size = 1 << 16
    grid_resolution = 128

    train_dataset = SubjectLoader(
        subject_id=args.scene,
        root_fp=args.data_root,
        split=args.train_split,
        num_rays=target_sample_batch_size // render_n_samples,
    )

    test_dataset = SubjectLoader(
        subject_id=args.scene,
        root_fp=args.data_root,
        split="test",
        num_rays=None,
    )

    occupancy_grid = OccupancyGrid(
        roi_aabb=args.aabb,
        resolution=grid_resolution,
        contraction_type=contraction_type,
    ).to(device)

    # training
    step = 0
    tic = time.time()
    for epoch in range(10000000):
        for i in range(len(train_dataset)):
            radiance_field.train()
            data = train_dataset[i]

            render_bkgd = data["color_bkgd"]
            rays = data["rays"]
            pixels = data["pixels"]
            timestamps = data["timestamps"]

            # update occupancy grid
            occupancy_grid.every_n_step(
                step=step,
                occ_eval_fn=lambda x: radiance_field.query_opacity(
                    x, timestamps, render_step_size
                ),
            )

            # render
            rgb, acc, depth, n_rendering_samples = render_image(
                radiance_field,
                occupancy_grid,
                rays,
                scene_aabb,
                # rendering options
                near_plane=near_plane,
                far_plane=far_plane,
                render_step_size=render_step_size,
                render_bkgd=render_bkgd,
                cone_angle=args.cone_angle,
                alpha_thre=0.01 if step > 1000 else 0.00,
                # dnerf options
                timestamps=timestamps,
            )
            if n_rendering_samples == 0:
                continue

            # dynamic batch size for rays to keep sample batch size constant.
            num_rays = len(pixels)
            num_rays = int(
                num_rays
                * (target_sample_batch_size / float(n_rendering_samples))
            )
            train_dataset.update_num_rays(num_rays)
            alive_ray_mask = acc.squeeze(-1) > 0

            # compute loss
            loss = F.smooth_l1_loss(rgb[alive_ray_mask], pixels[alive_ray_mask])

            optimizer.zero_grad()
            # do not unscale it because we are using Adam.
            grad_scaler.scale(loss).backward()
            optimizer.step()
            scheduler.step()

            if step % 5000 == 0:
                elapsed_time = time.time() - tic
                loss = F.mse_loss(rgb[alive_ray_mask], pixels[alive_ray_mask])
                print(
                    f"elapsed_time={elapsed_time:.2f}s | step={step} | "
                    f"loss={loss:.5f} | "
                    f"alive_ray_mask={alive_ray_mask.long().sum():d} | "
                    f"n_rendering_samples={n_rendering_samples:d} | num_rays={len(pixels):d} |"
                )

            if step > 0 and step % max_steps == 0:
                # evaluation
                radiance_field.eval()

                psnrs = []
                with torch.no_grad():
                    for i in tqdm.tqdm(range(len(test_dataset))):
                        data = test_dataset[i]
                        render_bkgd = data["color_bkgd"]
                        rays = data["rays"]
                        pixels = data["pixels"]
                        timestamps = data["timestamps"]

                        # rendering
                        rgb, acc, depth, _ = render_image(
                            radiance_field,
                            occupancy_grid,
                            rays,
                            scene_aabb,
                            # rendering options
                            near_plane=None,
                            far_plane=None,
                            render_step_size=render_step_size,
                            render_bkgd=render_bkgd,
                            cone_angle=args.cone_angle,
                            alpha_thre=0.01,
                            # test options
                            test_chunk_size=args.test_chunk_size,
                            # dnerf options
                            timestamps=timestamps,
                        )
                        mse = F.mse_loss(rgb, pixels)
                        psnr = -10.0 * torch.log(mse) / np.log(10.0)
                        psnrs.append(psnr.item())
                        # imageio.imwrite(
                        #     "acc_binary_test.png",
                        #     ((acc > 0).float().cpu().numpy() * 255).astype(np.uint8),
                        # )
                        # imageio.imwrite(
                        #     "rgb_test.png",
                        #     (rgb.cpu().numpy() * 255).astype(np.uint8),
                        # )
                        # break
                psnr_avg = sum(psnrs) / len(psnrs)
                print(f"evaluation: psnr_avg={psnr_avg}")
                train_dataset.training = True

            if step == max_steps:
                print("training stops")
                exit()

            step += 1