The Implementation For the FPS Measurement

[tingzhongyue]

Thank you for your great work and contribution.

I would like to ask about the code range for measuring FPS.

Did you measure the time of the render function or just the time of the rasterizer?

[Kevin-2017]

We render the time with the render function. Below are the code snippets. :)

def measure_fps(view, gaussians, pipe, background):
    t1_start = perf_counter() 
    render_pkg = render(view, gaussians, pipe, background)
    t1_stop = perf_counter()
    time = t1_stop - t1_start
    return time


def get_fps(dataset, opt, pipe, checkpoint, args):
    ic("fps")
    first_iter = 0
    tb_writer = prepare_output_and_logger(dataset)
    gaussians = GaussianModel(dataset.sh_degree)
    scene = Scene(dataset, gaussians)
    gaussians.training_setup(opt)
    if checkpoint:
        (model_params, _) = torch.load(checkpoint)
        gaussians.restore(model_params, opt)

    bg_color = [1, 1, 1] if dataset.white_background else [0, 0, 0]
    background = torch.tensor(bg_color, dtype=torch.float32, device="cuda")
    first_iter += 1
    test_camera = scene.getTestCameras().copy().pop()
    t_list = [0]*1000
    for i in range(1000):
        t_list[i] = measure_fps(test_camera, gaussians, pipe, background)
    t_list.sort()
    t_list = t_list[100:900]
    fps = 1 / t_list.mean
    ic(fps)