evaluate_tum.py

import cv2
import numpy as np
import glob
import os.path as osp
import os
import sys
import time
from pathlib import Path

import datetime
from tqdm import tqdm

from dpvo.utils import Timer
from dpvo.dpvo import DPVO
from dpvo.config import cfg
from dpvo.lietorch import SE3
import matplotlib.pyplot as plt
from imageio.v3 import imwrite
import torch
from multiprocessing import Process, Queue

### evo evaluation library ###
import evo
from evo.core.trajectory import PoseTrajectory3D
from evo.tools import file_interface
from evo.core import sync
import evo.main_ape as main_ape
from evo.core.metrics import PoseRelation
from evo.tools import plot
from dpvo.plot_utils import plot_trajectory, save_trajectory_tum_format


SKIP = 0

def show_image(image, t=0):
    image = image.permute(1, 2, 0).cpu().numpy()
    cv2.imshow('image', image / 255.0)
    cv2.waitKey(t)

def tum_image_stream(queue, scene_dir, sequence, stride, skip=0):
    """ image generator """
    images_dir = scene_dir / "rgb"

    fx, fy, cx, cy = 517.3, 516.5, 318.6, 255.3

    K_l = np.array([fx, 0.0, cx, 0.0, fy, cy, 0.0, 0.0, 1.0]).reshape(3,3)
    d_l = np.array([0.2624, -0.9531, -0.0054, 0.0026, 1.1633])

    image_list = sorted(images_dir.glob("*.png"))[skip::stride]

    for imfile in image_list:
        image = cv2.imread(str(imfile))
        image = cv2.undistort(image, K_l, d_l)
        image = cv2.resize(image, (320+32, 240+16))
        image = image.transpose(2,0,1)

        intrinsics = np.asarray([fx, fy, cx, cy])
        intrinsics[0] *= image.shape[2] / 640.0
        intrinsics[1] *= image.shape[1] / 480.0
        intrinsics[2] *= image.shape[2] / 640.0
        intrinsics[3] *= image.shape[1] / 480.0

        # crop image to remove distortion boundary
        intrinsics[2] -= 16
        intrinsics[3] -= 8
        # intrinsics = intrinsics[None]
        image = image[:, 8:-8, 16:-16]

        queue.put((float(imfile.stem), image, intrinsics))

    queue.put((-1, image, intrinsics))

@torch.no_grad()
def run(cfg, network, scene_dir, sequence, stride=1, viz=False):

    slam = None

    queue = Queue(maxsize=8)
    reader = Process(target=tum_image_stream, args=(queue, scene_dir, sequence, stride, 0))
    reader.start()

    for step in range(sys.maxsize):
        (t, images, intrinsics) = queue.get()
        if t < 0: break

        images = torch.as_tensor(images, device='cuda')
        intrinsics = torch.as_tensor(intrinsics, dtype=torch.float, device='cuda')

        if viz:
            show_image(images[0], 1)

        if slam is None:
            cam_poses = torch.as_tensor([-0.25, 0., 0., 0., 0., 0., 1.], dtype=torch.float, device='cuda')
            slam = DPVO(cfg, network, ht=images.shape[-2], wd=images.shape[-1], viz=viz)

        intrinsics = intrinsics.cuda()

        with Timer("SLAM", enabled=False):
            slam(t, images, intrinsics)

    for _ in range(12):
        slam.update()

    reader.join()

    poses, tstamps = slam.terminate()
    np.save(f"poses_{sequence}.npy", poses)
    np.save(f"tstamps_{sequence}.npy", tstamps)
    return poses, tstamps


if __name__ == '__main__':
    import argparse
    parser = argparse.ArgumentParser()
    parser.add_argument('--network', type=str, default='dpvo.pth')
    parser.add_argument('--config', default="config/default.yaml")
    parser.add_argument('--stride', type=int, default=2)
    parser.add_argument('--viz', action="store_true")
    parser.add_argument('--trials', type=int, default=1)
    parser.add_argument('--tumdir', type=Path, default="datasets/TUM_RGBD")
    parser.add_argument('--plot', action="store_true")
    parser.add_argument('--save_trajectory', action="store_true")
    args = parser.parse_args()

    cfg.merge_from_file(args.config)

    print("\nRunning with config...")
    print(cfg, "\n")

    main_seed = int(time.time())
    print(f"main_seed: {main_seed}")
    torch.manual_seed(main_seed)

    tum_scenes = [
        "360",
        "desk",
        "desk2",
        "floor",
        "plant",
        "room",
        "rpy",
        "teddy",
        "xyz",
    ]

    results = {}
    for scene in tum_scenes:
        scene_dir = args.tumdir / "frieburg1" / f"rgbd_dataset_freiburg1_{scene}"
        groundtruth = scene_dir / "groundtruth.txt"#"dataset" / "poses" / f"{scene}.txt"
        traj_ref = file_interface.read_tum_trajectory_file(groundtruth)

        scene_results = []
        for trial_num in range(args.trials):
            traj_est, timestamps = run(cfg, args.network, scene_dir, scene, args.stride, args.viz)

            traj_est = PoseTrajectory3D(
                positions_xyz=traj_est[:,:3],
                orientations_quat_wxyz=traj_est[:,3:],
                timestamps=timestamps)

            traj_ref, traj_est = sync.associate_trajectories(traj_ref, traj_est)

            result = main_ape.ape(traj_ref, traj_est, est_name='traj', 
                pose_relation=PoseRelation.translation_part, align=True, correct_scale=True)
            ate_score = result.stats["rmse"]

            if args.plot:
                Path("trajectory_plots").mkdir(exist_ok=True)
                plot_trajectory(traj_est, traj_ref, f"TUM-RGBD Frieburg1 {scene} Trial #{trial_num+1} (ATE: {ate_score:.03f})",
                                f"trajectory_plots/TUM_RGBD_Frieburg1_{scene}_Trial{trial_num+1:02d}.pdf", align=True, correct_scale=True)

            if args.save_trajectory:
                Path("saved_trajectories").mkdir(exist_ok=True)
                save_trajectory_tum_format(traj_est, f"saved_trajectories/TUM_RGBD_{scene}_Trial{trial_num+1:02d}.txt")

            scene_results.append(ate_score)

        results[scene] = np.median(scene_results)
        print(scene, sorted(scene_results))

    xs = []
    for scene in results:
        print(scene, results[scene])
        xs.append(results[scene])

    print("AVG: ", np.mean(xs))