open3D 11

[신민서]

11. OpenPCDet

예측 확률이 0.5 이상인 객체만 바운딩 박스를 그려라.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

from inference2 import *   if __name__ == '__main__':       cfg_file = "pointpillar.yaml"     #cfg_file = "pointrcnn.yaml"     #cfg_file = "second.yaml"     #cfg_file = "voxel_rcnn_car.yaml"     #cfg_file = "pv_rcnn.yaml"       data_path = "000000.bin"     #data_path = "000008.bin"              ckpt_path = "pointpillar_7728.pth"     #ckpt_path = "pointrcnn_7870.pth"     #ckpt_path = "second_7862.pth"     #ckpt_path = "voxel_rcnn_car_84.54.pth"     #ckpt_path = "pv_rcnn_8369.pth"          pred_dict, pcd = model_inference(cfg_file=cfg_file, data_path=data_path, ckpt_path=ckpt_path)     print(pred_dict['pred_scores'])     vis = open3d.visualization.Visualizer()     vis.create_window(window_name="kitti")     vis.get_render_option().point_size = 2.0     vis.get_render_option().background_color = np.zeros(3)     axis_pcd = open3d.geometry.TriangleMesh.create_coordinate_frame(size=1.0, origin=[0, 0, 0])     vis.add_geometry(axis_pcd)       pts = open3d.geometry.PointCloud()     pts.points = open3d.utility.Vector3dVector(pcd)     pts.colors = open3d.utility.Vector3dVector(np.ones((len(pcd), 3)))     vis.add_geometry(pts)       box_color = [[0,1,0],[0,1,1],[1,1,0],[1,1,1]]       for i in range(len(pred_dict['pred_boxes'])):         if pred_dict['pred_scores'][i] > 0.5: # 0.5 이상만 바운딩 박스             print(pred_dict['pred_scores'][i]) # print로 결과 출력             center = pred_dict['pred_boxes'][i][0:3]             lwh = pred_dict['pred_boxes'][i][3:6]             axis_angles = np.array([0, 0, pred_dict['pred_boxes'][i][6]])             rot = open3d.geometry.get_rotation_matrix_from_axis_angle(axis_angles)             box3d = open3d.geometry.OrientedBoundingBox(center, rot, lwh)             box3d.color = box_color[pred_dict['pred_labels'][i]-1]              vis.add_geometry(box3d)                         vis.run()     vis.destroy_window() Colored by Color Scripter

cs

레이블 파일에 있는 3d 바운딩 박스 정보를 예측 결과와 같이 그려보라,

색상은 달리하여 정답과 예측 결과가 얼마나 일치하는지 확인하라.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57

from open3d_func import * from inference2 import *   if __name__ == '__main__':       cfg_file = "pointpillar.yaml"     #cfg_file = "pointrcnn.yaml"     #cfg_file = "second.yaml"     #cfg_file = "voxel_rcnn_car.yaml"     #cfg_file = "pv_rcnn.yaml"       data_path = "000000.bin"     #data_path = "000008.bin"              ckpt_path = "pointpillar_7728.pth"     #ckpt_path = "pointrcnn_7870.pth"     #ckpt_path = "second_7862.pth"     #ckpt_path = "voxel_rcnn_car_84.54.pth"     #ckpt_path = "pv_rcnn_8369.pth"       calib_file_path = "../OpenPCDet/data/kitti/training/calib/000000.txt"     label_file_path = "../OpenPCDet/data/kitti/training/label_2/000000.txt"          pred_dict,pcd = model_inference(cfg_file=cfg_file, data_path=data_path, ckpt_path=ckpt_path)          vis = open3d.visualization.Visualizer()     vis.create_window(window_name="kitti")     vis.get_render_option().point_size = 2.0     vis.get_render_option().background_color = np.zeros(3)     axis_pcd = open3d.geometry.TriangleMesh.create_coordinate_frame(size=1.0, origin=[0, 0, 0])     vis.add_geometry(axis_pcd)       pts = open3d.geometry.PointCloud()     pts.points = open3d.utility.Vector3dVector(pcd)     pts.colors = open3d.utility.Vector3dVector(np.ones((len(pcd), 3)))     vis.add_geometry(pts)       calib = parse_kitti_calibration_file(calib_file_path)     labels = parse_kitti_label_file(label_file_path)     pcd = load_point_cloud(data_path)       calib = parse_kitti_calibration_file(calib_file_path)     labels = parse_kitti_label_file(label_file_path)     gt_bboxes = []     for label in labels:         if label["type"] == "DontCare":             continue         gt_bboxes.append(open3d_bbox(label, calib))       # 시각화 (정답 + 예측 같이)     visualize_pcd(         pcd=pcd,         gt_bboxes=gt_bboxes,         pred_boxes=pred_dict["pred_boxes"],         pred_labels=pred_dict["pred_labels"],         label_colors=[[0, 1, 0]] * len(pred_dict["pred_boxes"])  # 모든 박스 초록색 ) Colored by Color Scripter

cs

예측: 초록색 ///// 레이블: 빨간색

보행자만 label 해놨기 때문에 실제로 하나가 맞음

추론 결과를 이용하여 영상 파일에 3d 바운딩 박스를 그려라,

먼저 라이다 좌표계에서 8개의 3차원 꼭지점 좌표(get_box_points 함수 이용)를 구하고

좌표변환 행렬(Tr_vel_to_cam -> R0_rect -> P2)을 이용하여 영상 좌표계의 8개의 2차원 꼭지점으로 변환

opencv 함수를 이용하여 라인을 그려라.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49

from inference2 import * from open3d_func import *   if __name__ == '__main__':       cfg_file = "pointpillar.yaml"     #cfg_file = "pointrcnn.yaml"     #cfg_file = "second.yaml"     #cfg_file = "voxel_rcnn_car.yaml"     #cfg_file = "pv_rcnn.yaml"       data_path = "000000.bin"     #data_path = "000008.bin"              ckpt_path = "pointpillar_7728.pth"     #ckpt_path = "pointrcnn_7870.pth"     #ckpt_path = "second_7862.pth"     #ckpt_path = "voxel_rcnn_car_84.54.pth"     #ckpt_path = "pv_rcnn_8369.pth"       calib_file_path = "../OpenPCDet/data/kitti/training/calib/000000.txt"     label_file_path = "../OpenPCDet/data/kitti/training/label_2/000000.txt"          pred_dict,pcd = model_inference(cfg_file=cfg_file, data_path=data_path, ckpt_path=ckpt_path)       calib = parse_kitti_calibration_file(calib_file_path)     labels = parse_kitti_label_file(label_file_path)     pcd = load_point_cloud(data_path)       calib = parse_kitti_calibration_file(calib_file_path)     labels = parse_kitti_label_file(label_file_path)       gt_bboxes = []       for label in labels:         if label["type"] == "DontCare":             continue         gt_bboxes.append(open3d_bbox(label, calib))       image = draw_predicted_boxes_on_image(         image_path="../OpenPCDet/data/kitti/training/image_2/000000.png",         calib_path=calib_file_path,         pred_boxes=pred_dict["pred_boxes"],         color=(0, 255, 0)     )       cv2.imshow("Prediction", image)     cv2.waitKey(0)     cv2.destroyAllWindows() Colored by Color Scripter

cs