import numpy as np import os import trimesh import torch import argparse from ChamferDistancePytorch.chamfer3D import dist_chamfer_3D def load_trimesh_df3d(root_dir): folders = os.listdir(root_dir) meshs = {} for i, f in enumerate(folders): sub_name = f meshs[sub_name] = trimesh.load(os.path.join(root_dir, f, '%s.obj' % (sub_name+'.obj_scaled'))) return meshs if __name__ == '__main__': parser = argparse.ArgumentParser( description='Run error computing' ) parser.add_argument('--results_path', type=str, default='./results_eval', help='path to save results') parser.add_argument('--root_path', type=str, help='path to dataset root') args = parser.parse_args() root_path = args.root_path OBJ_path = os.path.join(root_path, 'GEO', 'OBJ') PARAM_path = os.path.join(root_path, 'PARAM') chamL2 = dist_chamfer_3D.chamfer_3DDist() var_subjects = list(np.loadtxt(os.path.join(root_path, 'val.txt'), dtype=str)) yaw_list = [0, 40, 320] chamL2_log = os.path.join(args.results_path, 'chamL2_p2s_log.txt') dist_mean_list = [] p2s_mean_list = [] for subject in var_subjects: print(subject) with open(chamL2_log, "a") as log_file: log_file.write('%s\n' % subject) dist_pc_list = [] dist_p2s_list = [] result_npz_path = os.path.join(args.results_path, 'npz', subject) mesh = trimesh.load(os.path.join(OBJ_path, subject, '%s.obj' % (subject+'.obj_scaled'))) sample_point_count = 1000000 surface_point_cloud = mesh.sample(sample_point_count, return_index=False) surface_point_cloud = torch.Tensor(surface_point_cloud).float().cuda() for vid in yaw_list: npz_file = os.path.join(result_npz_path, '%d_%d_%02d.npz' % (vid, 0, 0)) result_npz = np.load(npz_file) result_points = result_npz['points'] result_points = torch.Tensor(result_points).unsqueeze(0).float().cuda() result_points = (result_points + 1) / 2.0 param_path = os.path.join(PARAM_path, subject, '%d_%d_%02d.npy' % (vid, 0, 0)) param = np.load(param_path, allow_pickle=True) ortho_ratio = param.item().get('ortho_ratio') scale = param.item().get('scale') center = param.item().get('center') R = param.item().get('R') translate = -np.matmul(R, center).reshape(3, 1) extrinsic = np.concatenate([R, translate], axis=1) extrinsic = np.concatenate([extrinsic, np.array([0, 0, 0, 1]).reshape(1, 4)], 0) scale_intrinsic = np.identity(4) scale_intrinsic[0, 0] = scale / ortho_ratio scale_intrinsic[1, 1] = scale / ortho_ratio scale_intrinsic[2, 2] = scale / ortho_ratio uv_intrinsic = np.identity(4) uv_intrinsic[0, 0] = 1.0 / float(512 // 2) uv_intrinsic[1, 1] = 1.0 / float(512 // 2) uv_intrinsic[2, 2] = 1.0 / float(512 // 2) trans_intrinsic = np.identity(4) intrinsic = np.matmul(trans_intrinsic, np.matmul(uv_intrinsic, scale_intrinsic)) calib = torch.Tensor(np.matmul(intrinsic, extrinsic)).float().cuda() rot = calib[:3, :3] trans = calib[:3, 3:4] proj_gt = torch.matmul(rot, surface_point_cloud.T) + trans proj_gt = proj_gt.T.unsqueeze(0) proj_gt = (proj_gt + 1) / 2.0 with torch.no_grad(): dist_pc1, dist_pc2, _, _ = chamL2(result_points, proj_gt) # error_pc1 = [B, M], error_pc2 = [B, M] dist_pc = (dist_pc1.mean() + dist_pc2.mean()) * 0.5 dist_pc_list.append(dist_pc) dist_p2s_list.append(dist_pc1.mean()) dist_mean = torch.stack(dist_pc_list).mean() dist_mean_list.append(dist_mean) p2s_mean = torch.stack(dist_p2s_list).mean() p2s_mean_list.append(p2s_mean) message = 'Chamer L2: {}'.format(dist_mean.item()) print(message) with open(chamL2_log, "a") as log_file: log_file.write('%s\n' % message) message = 'P2S L2: {}'.format(p2s_mean.item()) print(message) with open(chamL2_log, "a") as log_file: log_file.write('%s\n' % message) dist_final = torch.stack(dist_mean_list).mean() p2s_final = torch.stack(p2s_mean_list).mean() message = 'Final Chamer L2: {}'.format(dist_final.item()) print(message) with open(chamL2_log, "a") as log_file: log_file.write('%s\n' % message) message = 'Final P2S L2: {}'.format(p2s_final.item()) print(message) with open(chamL2_log, "a") as log_file: log_file.write('%s\n' % message)