3D Gaussian Splatting - FPS 측정하기

3D-GS 계열 전체에 걸쳐 쓸 수 있는 FPS 측정 스크립트를 야매로 구성해봤다. EAGLES의 코드를 많이 참고함.

사용법은 3dgs 코드의 render.py와 거의 유사하다

 

아래 소스코드를 복사한 후, 3d-gs 디렉토리에 새로운 python 파일 하나 붙여넣은 다음에 사용하면 된다. 나는 measure_fps.py로 이름 짓고 사용 중.

import os
import json
import torch
import numpy as np
import subprocess as sp

from gaussian_renderer import render
import torch.utils.benchmark as benchmark

from gaussian_renderer import GaussianModel
from argparse import ArgumentParser
from arguments import ModelParams, PipelineParams, get_combined_args
from scene import Scene

def render_fn(views, gaussians, pipeline, background):
    with torch.autocast(device_type='cuda', dtype=torch.float16):
        for view in views:
            render(view, gaussians, pipeline, background)

def measure_fps(scene, gaussians, pipeline, background):
    with torch.no_grad():
        views = scene.getTrainCameras() + scene.getTestCameras()
        t0 = benchmark.Timer(stmt='render_fn(views, gaussians, pipeline, background)',
                            setup='from __main__ import render_fn',
                            globals={'views': views, 'gaussians': gaussians, 'pipeline': pipeline, 
                                     'background': background},
                            )
        time = t0.timeit(50)
        fps = len(views)/time.median
        print("Rendering FPS: ", fps)
    return fps

def render_sets(dataset : ModelParams, iteration : int, pipeline : PipelineParams):
    with torch.no_grad():
        gaussians = GaussianModel(dataset.sh_degree)
        scene = Scene(dataset, gaussians, load_iteration=iteration, shuffle=False)

        bg_color = [1,1,1] if dataset.white_background else [0, 0, 0]
        background = torch.tensor(bg_color, dtype=torch.float32, device="cuda")

        fps = measure_fps(scene, gaussians, pipeline, background)

if __name__ == "__main__":
    # Set up command line argument parser
    parser = ArgumentParser(description="Testing script parameters")
    model = ModelParams(parser, sentinel=True)
    pipeline = PipelineParams(parser)
    parser.add_argument("--iteration", default=-1, type=int)
    args = get_combined_args(parser)
    print("Rendering " + args.model_path)

    # model.extract(args), args.iteration, pipeline.extract(args), args.skip_train, args.skip_test
    print(f"model_extract: {model.extract(args)}")
    print(f"args_iteration: {args.iteration}")
    print(f"pipeline_extract: {pipeline.extract(args)}")
    # print(f"args_skip_train: {args.skip_train}")
    # print(f"args_skip_test: {args.skip_test}")
    render_sets(model.extract(args), args.iteration, pipeline.extract(args))

 

실행 예시는 아래와 같다.

python measure_fps.py -m output/fern -s data/fern