def main(): import argparse print('Parsing the arguments...') parser = argparse.ArgumentParser() # exp name # parser.add_argument( '--exp_name', required=True, type=str, help='The name of the experiment. Used to naming the folders') # about testing # parser.add_argument('--img_name_pattern', type=str, default='*.png', help='image name pattern') parser.add_argument('--model_path', type=str, default='.', help='The pre-trained model path for KV-net') parser.add_argument('--split_file', type=str, default='.', help='The split txt file') parser.add_argument('--t_win', type=int, default=2, help='The radius of the temporal window; default=2') parser.add_argument('--d_min', type=float, default=0, help='The minimal depth value; default=0') parser.add_argument('--d_max', type=float, default=5, help='The maximal depth value; default=15') parser.add_argument('--ndepth', type=int, default=64, help='The # of candidate depth values; default= 128') parser.add_argument('--sigma_soft_max', type=float, default=10., help='sigma_soft_max, default = 500.') parser.add_argument( '--feature_dim', type=int, default=64, help='The feature dimension for the feature extractor; default=64') # about pose # parser.add_argument('--intrin_path', type=str, required=True, help='camera intrinic path, saved as .mat') parser.add_argument( '--dso_res_path', type=str, default='dso_res/result_dso.txt', help= 'if use DSO pose, specify the path to the DSO results. Should be a .txt file' ) parser.add_argument('--opt_next_frame', action='store_true', help='') parser.add_argument('--use_gt_R', action='store_true', help='') parser.add_argument('--use_gt_t', action='store_true', help='') parser.add_argument('--use_dso_R', action='store_true', help='') parser.add_argument('--use_dso_t', action='store_true', help='') parser.add_argument('--min_frame_idx', type=int, help=' ', default=0) parser.add_argument('--max_frame_idx', type=int, help=' ', default=10000) parser.add_argument('--refresh_frames', type=int, help=' ', default=1000) parser.add_argument('--LBA_max_iter', type=int, help=' ') parser.add_argument('--opt_r', type=int, default=1, help=' ') parser.add_argument('--opt_t', type=int, default=1, help=' ') parser.add_argument('--LBA_step', type=float, help=' ') parser.add_argument('--frame_interv', type=int, default=5, help=' ') # about dataset # parser.add_argument('--dataset', type=str, default='7scenes', help='Dataset name: {scanNet, 7scenes}') parser.add_argument('--dataset_path', type=str, default='.', help='Path to the dataset') # about output # parser.add_argument('--output_pngs', action='store_true', help='if output pngs') # para config. # args = parser.parse_args() exp_name = args.exp_name dataset_name = args.dataset t_win_r = args.t_win nDepth = args.ndepth d_candi = np.linspace(args.d_min, args.d_max, nDepth) sigma_soft_max = args.sigma_soft_max #10.#500. dnet_feature_dim = args.feature_dim frame_interv = args.frame_interv d_candi_dmap_ref = d_candi nDepth_dmap_ref = nDepth # Initialize data-loader, model and optimizer # # ===== Dataset selection ======== # dataset_path = args.dataset_path if dataset_name == 'scanNet': # deal with 1-frame scanNet data import mdataloader.scanNet as dl_scanNet dataset_init = dl_scanNet.ScanNet_dataset split_txt = './mdataloader/scanNet_split/scannet_val.txt' if args.split_file == '.' else args.split_file if not dataset_path == '.': # if specify the path, we will assume we are using 1-frame-interval scanNet video # fun_get_paths = lambda traj_indx: dl_scanNet.get_paths_1frame( traj_indx, database_path_base=dataset_path, split_txt=split_txt) dat_indx_step = 5 #pick this value to make sure the camera baseline is big enough else: fun_get_paths = lambda traj_indx: dl_scanNet.get_paths( traj_indx, frame_interv=5, split_txt=split_txt) dat_indx_step = 1 img_size = [384, 256] # trajectory index for training # n_scenes, _, _, _, _ = fun_get_paths(0) traj_Indx = np.arange(0, n_scenes) elif dataset_name == '7scenes': # 7 scenes video # import mdataloader.dl_7scenes as dl_7scenes img_size = [384, 256] dataset_init = dl_7scenes.SevenScenesDataset dat_indx_step = 5 # pick this value to make sure the camera baseline is big enough # trajectory index for training # split_file = None if args.split_file == '.' else args.split_file fun_get_paths = lambda traj_indx: dl_7scenes.get_paths_1frame( traj_indx, database_path_base=dataset_path, split_txt=split_file, ) elif dataset_name == 'single_folder': # images in a single folder specified by the user # import mdataloader.mdata as mdata img_size = [384, 256] dataset_init = mdata.mData dat_indx_step = 5 # pick this value to make sure the camera baseline is big enough fun_get_paths = lambda traj_indx: mdata.get_paths_1frame( traj_indx, dataset_path, args.img_name_pattern) traj_Indx = [0] #dummy fldr_path, img_paths, dmap_paths, poses, intrin_path = fun_get_paths( traj_Indx[0]) if dataset_name == 'single_folder': intrin_path = args.intrin_path dataset = dataset_init( True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, resize_dmap=.25, ) dataset_Himgsize = dataset_init( True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, resize_dmap=.5, ) dataset_imgsize = dataset_init( True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, resize_dmap=1, ) # ================================ # print('Initnializing the KV-Net') model_KVnet = m_kvnet.KVNET(\ feature_dim = dnet_feature_dim, cam_intrinsics = dataset.cam_intrinsics, d_candi = d_candi, sigma_soft_max = sigma_soft_max, KVNet_feature_dim = dnet_feature_dim, d_upsample_ratio_KV_net = None, t_win_r = t_win_r, if_refined = True) model_KVnet = torch.nn.DataParallel(model_KVnet) model_KVnet.cuda() model_path_KV = args.model_path print('loading KV_net at %s' % (model_path_KV)) utils_model.load_pretrained_model(model_KVnet, model_path_KV) print('Done') for traj_idx in traj_Indx: scene_path_info = [] print('Getting the paths for traj_%d' % (traj_idx)) fldr_path, img_seq_paths, dmap_seq_paths, poses, intrin_path = fun_get_paths( traj_idx) res_fldr = '../results/%s/traj_%d' % (exp_name, traj_idx) m_misc.m_makedir(res_fldr) dataset.set_paths(img_seq_paths, dmap_seq_paths, poses) if dataset_name == 'scanNet': # the camera intrinsic may be slightly different for different trajectories in scanNet # dataset.get_cam_intrinsics(intrin_path) print('Done') if args.min_frame_idx > 0: frame_idxs = np.arange(args.min_frame_idx - t_win_r, args.min_frame_idx + t_win_r) dat_array = [dataset[idx] for idx in frame_idxs] else: dat_array = [dataset[idx] for idx in range(t_win_r * 2 + 1)] DMaps_meas = [] dso_res_path = args.dso_res_path print('init initial pose from DSO estimations ...') traj_extMs = init_traj_extMs(traj_len=len(dataset), dso_res_path=dso_res_path, if_filter=True, min_idx=args.min_frame_idx, max_idx=args.max_frame_idx) traj_extMs_init = copy_list(traj_extMs) traj_length = min(len(dataset), len(traj_extMs)) first_frame = True for frame_cnt, ref_indx in enumerate( range(t_win_r * dat_indx_step + args.min_frame_idx, traj_length - t_win_r * dat_indx_step - dat_indx_step)): # ref_indx: the frame index for the reference frame # # Read ref. and src. data in the local time window # ref_dat, src_dats = m_misc.split_frame_list(dat_array, t_win_r) src_frame_idx = [ idx for idx in range( ref_indx - t_win_r * dat_indx_step, ref_indx, dat_indx_step) ] + \ [ idx for idx in range( ref_indx + dat_indx_step, ref_indx + t_win_r*dat_indx_step+1, dat_indx_step) ] valid_seq = dso_io.valid_poses(traj_extMs, src_frame_idx) # only look at a subset of frames # if ref_indx < args.min_frame_idx: valid_seq = False if ref_indx > args.max_frame_idx or ref_indx >= traj_length - t_win_r * dat_indx_step - dat_indx_step: break if frame_cnt == 0 or valid_seq is False: BVs_predict = None # refresh # if ref_indx % args.refresh_frames == 0: print('REFRESH !') BVs_predict = None BVs_predict_in = None first_frame = True traj_extMs = copy_list(traj_extMs_init) if valid_seq: # if the sequence does not contain invalid pose estimation # Get poses # src_cam_extMs = [traj_extMs[i] for i in src_frame_idx] ref_cam_extM = traj_extMs[ref_indx] src_cam_poses = [ warp_homo.get_rel_extrinsicM(ref_cam_extM, src_cam_extM_) for src_cam_extM_ in src_cam_extMs ] src_cam_poses = [ torch.from_numpy(pose.astype( np.float32)).cuda().unsqueeze(0) for pose in src_cam_poses ] # Load the gt pose if available # if 'extM' in dataset[0]: src_cam_extMs_ref = [ dataset[i]['extM'] for i in src_frame_idx ] ref_cam_extM_ref = dataset[ref_indx]['extM'] src_cam_poses_ref = [ warp_homo.get_rel_extrinsicM(ref_cam_extM_ref, src_cam_extM_) \ for src_cam_extM_ in src_cam_extMs_ref ] src_cam_poses_ref = [ torch.from_numpy(pose.astype(np.float32)).cuda().unsqueeze(0) \ for pose in src_cam_poses_ref ] # -- Determine the scale, mapping from DSO scale to our working scale -- # if frame_cnt == 0 or BVs_predict is None: # the first window for the traj. _, t_norm_single = get_fb(src_cam_poses, dataset.cam_intrinsics, src_cam_pose_next=None) # We need to heurisitcally determine scale_ without using GT pose # t_norms = get_t_norms(traj_extMs, dat_indx_step) scale_ = d_candi.max() / ( dataset.cam_intrinsics['focal_length'] * np.array(t_norm_single).max() / 2) scale_ = d_candi.max() / ( dataset.cam_intrinsics['focal_length'] * np.array(t_norms).max()) scale_ = d_candi.max() / ( dataset.cam_intrinsics['focal_length'] * np.array(t_norms).mean() / 2) rescale_traj_t(traj_extMs, scale_) traj_extMs_dso = copy_list(traj_extMs) # Get poses # src_cam_extMs = [traj_extMs[i] for i in src_frame_idx] ref_cam_extM = traj_extMs[ref_indx] src_cam_poses = [ warp_homo.get_rel_extrinsicM(ref_cam_extM, src_cam_extM_) for src_cam_extM_ in src_cam_extMs ] src_cam_poses = [ torch.from_numpy(pose.astype( np.float32)).cuda().unsqueeze(0) for pose in src_cam_poses ] # src_cam_poses size: N V 4 4 # src_cam_poses = torch.cat(src_cam_poses, dim=0).unsqueeze(0) src_frames = [m_misc.get_entries_list_dict(src_dats, 'img')] cam_pose_next = traj_extMs[ref_indx + 1] cam_pose_next = torch.FloatTensor( warp_homo.get_rel_extrinsicM(traj_extMs[ref_indx], cam_pose_next)).cuda() BVs_predict_in = None if frame_cnt == 0 or BVs_predict is None \ else BVs_predict BVs_measure, BVs_predict = test_KVNet.test( model_KVnet, d_candi, Ref_Dats=[ref_dat], Src_Dats=[src_dats], Cam_Intrinsics=[dataset.cam_intrinsics], t_win_r=t_win_r, Src_CamPoses=src_cam_poses, BV_predict=BVs_predict_in, R_net=True, cam_pose_next=cam_pose_next, ref_indx=ref_indx) # export_res.export_res_refineNet(ref_dat, BVs_measure, d_candi_dmap_ref, # res_fldr, ref_indx, # save_mat = True, output_pngs = args.output_pngs, output_dmap_ref=False) export_res.export_res_img(ref_dat, BVs_measure, d_candi_dmap_ref, res_fldr, frame_cnt) scene_path_info.append( [frame_cnt, dataset[ref_indx]['img_path']]) # UPDATE dat_array # if dat_indx_step > 1: # use one-interval video and the frame interval is larger than 5 print('updating array ...') dat_array = update_dat_array(dat_array, dataset, data_interv=1, frame_interv=5, ref_indx=ref_indx, t_win_r=t_win_r) print('done') else: dat_array.pop(0) new_dat = dataset[ref_indx + t_win_r + 1] dat_array.append(new_dat) # OPTMIZE POSES # idx_ref_ = ref_indx + 1 cam_pose_nextframe = traj_extMs[idx_ref_] cam_pose_nextframe = torch.FloatTensor( warp_homo.get_rel_extrinsicM(traj_extMs[ref_indx], cam_pose_nextframe)).cuda() # get depth and confidence map # BV_tmp_ = warp_homo.resample_vol_cuda(\ src_vol = BVs_measure, rel_extM = cam_pose_nextframe.inverse(), cam_intrinsic = dataset_imgsize.cam_intrinsics, d_candi = d_candi, d_candi_new = d_candi, padding_value = math.log(1. / float(len(d_candi))) ).clamp(max=0, min=-1000.) dmap_ref = m_misc.depth_val_regression(BVs_measure, d_candi, BV_log=True).squeeze() conf_map_ref, _ = torch.max(BVs_measure.squeeze(), dim=0) dmap_kf = m_misc.depth_val_regression(BV_tmp_.unsqueeze(0), d_candi, BV_log=True).squeeze() conf_map_kf, _ = torch.max(BV_tmp_.squeeze(), dim=0) # setup optimization # cams_intrin = [ dataset.cam_intrinsics, dataset_Himgsize.cam_intrinsics, dataset_imgsize.cam_intrinsics ] dw_scales = [4, 2, 1] LBA_max_iter = args.LBA_max_iter #10 # 20 LBA_step = args.LBA_step #.05 #.01 if LBA_max_iter <= 1: # do not do optimization LBA_step = 0. opt_vars = [args.opt_r, args.opt_t] # initialization for the first time window # if first_frame: first_frame = False # optimize the pose for all frames within the window # if LBA_max_iter <= 1: # for debugging: using GT pose initialization # rel_pose_inits_all_frame, srcs_idx_all_frame = m_misc.get_twin_rel_pose( traj_extMs, idx_ref_, t_win_r * dat_indx_step, 1, use_gt_R=True, use_gt_t=True, dataset=dataset, add_noise_gt=False, noise_sigmas=None) else: rel_pose_inits_all_frame, srcs_idx_all_frame = m_misc.get_twin_rel_pose( traj_extMs, ref_indx, t_win_r * dat_indx_step, 1, use_gt_R=False, use_gt_t=False, dataset=dataset, ) # opt. # img_ref = dataset[ref_indx]['img'] imgs_src = [dataset[i]['img'] for i in srcs_idx_all_frame] conf_map_ref = torch.exp(conf_map_ref).squeeze()**2 rel_pose_opt = opt_pose_numerical.local_BA_direct( img_ref, imgs_src, dmap_ref.unsqueeze(0).unsqueeze(0), conf_map_ref.unsqueeze(0).unsqueeze(0), cams_intrin, dw_scales, rel_pose_inits_all_frame, max_iter=LBA_max_iter, step=LBA_step, opt_vars=opt_vars) # update # for idx, srcidx in enumerate(srcs_idx_all_frame): traj_extMs[srcidx] = np.matmul( rel_pose_opt[idx].cpu().numpy(), traj_extMs[ref_indx]) # for next frame # if LBA_max_iter <= 1: # for debugging: using GT pose init. rel_pose_opt, srcs_idx = m_misc.get_twin_rel_pose( traj_extMs, idx_ref_, t_win_r, dat_indx_step, use_gt_R=True, use_gt_t=True, dataset=dataset, add_noise_gt=False, noise_sigmas=None, ) else: rel_pose_inits, srcs_idx = m_misc.get_twin_rel_pose( traj_extMs, idx_ref_, t_win_r, dat_indx_step, use_gt_R=args.use_gt_R, use_dso_R=args.use_dso_R, use_gt_t=args.use_gt_t, use_dso_t=args.use_dso_t, dataset=dataset, traj_extMs_dso=traj_extMs_dso, opt_next_frame=args.opt_next_frame) img_ref = dataset[idx_ref_]['img'] _, src_dats_opt = m_misc.split_frame_list( dat_array, t_win_r) imgs_src = [dat_['img'] for dat_ in src_dats_opt] img_ref = dataset[idx_ref_]['img'] imgs_src = [dataset[i] for i in srcs_idx] imgs_src = [img_['img'] for img_ in imgs_src] # opt. # conf_map_kf = torch.exp(conf_map_kf).squeeze()**2 rel_pose_opt = \ opt_pose_numerical.local_BA_direct_parallel( img_ref, imgs_src, dmap_kf.unsqueeze(0).unsqueeze(0), conf_map_kf.unsqueeze(0).unsqueeze(0), cams_intrin, dw_scales, rel_pose_inits, max_iter = LBA_max_iter, step = LBA_step, opt_vars = opt_vars) # update # print('idx_ref_: %d' % (idx_ref_)) print('srcs_idx : ') print(srcs_idx) print('updating pose ...') for idx, srcidx in enumerate(srcs_idx): traj_extMs[srcidx] = np.matmul( rel_pose_opt[idx].cpu().numpy(), traj_extMs[idx_ref_]) print('done') else: # if the sequence contains invalid pose estimation BVs_predict = None print('frame_cnt :%d, include invalid poses' % (frame_cnt)) # UPDATE dat_array # if dat_indx_step > 1: # use one-interval video and the frame interval is larger than 5 print('updating array ...') dat_array = update_dat_array(dat_array, dataset, data_interv=1, frame_interv=5, ref_indx=ref_indx, t_win_r=t_win_r) print('done') else: dat_array.pop(0) new_dat = dataset[ref_indx + t_win_r + 1] dat_array.append(new_dat) m_misc.save_ScenePathInfo('%s/scene_path_info.txt' % (res_fldr), scene_path_info)
def main(): import argparse print('Parsing the arguments...') parser = argparse.ArgumentParser() # exp name # parser.add_argument( '--exp_name', required=True, type=str, help='The name of the experiment. Used to naming the folders') # about testing # parser.add_argument('--model_path', type=str, required=True, help='The pre-trained model path for KV-net') parser.add_argument('--split_file', type=str, required=True, help='The split txt file') parser.add_argument('--frame_interv', default=5, type=int, help='frame interval') parser.add_argument('--t_win', type=int, default=2, help='The radius of the temporal window; default=2') parser.add_argument('--d_min', type=float, default=0, help='The minimal depth value; default=0') parser.add_argument('--d_max', type=float, default=5, help='The maximal depth value; default=15') parser.add_argument('--ndepth', type=int, default=64, help='The # of candidate depth values; default= 128') parser.add_argument('--sigma_soft_max', type=float, default=10., help='sigma_soft_max, default = 500.') parser.add_argument( '--feature_dim', type=int, default=64, help='The feature dimension for the feature extractor; default=64') # about dataset # parser.add_argument('--dataset', type=str, default='scanNet', help='Dataset name: {scanNet, 7scenes, kitti}') parser.add_argument('--dataset_path', type=str, default='.', help='Path to the dataset') parser.add_argument( '--change_aspect_ratio', action='store_true', default=False, help= 'If we want to change the aspect ratio. This option is only useful for KITTI' ) # parsing parameters # args = parser.parse_args() exp_name = args.exp_name dataset_name = args.dataset t_win_r = args.t_win nDepth = args.ndepth d_candi = np.linspace(args.d_min, args.d_max, nDepth) sigma_soft_max = args.sigma_soft_max #10.#500. dnet_feature_dim = args.feature_dim frame_interv = args.frame_interv # should be multiple of 5 for scanNet dataset d_upsample = None d_candi_dmap_ref = d_candi nDepth_dmap_ref = nDepth # ===== Dataset selection ======== # dataset_path = args.dataset_path if dataset_name == 'scanNet': import mdataloader.scanNet as dl_scanNet dataset_init = dl_scanNet.ScanNet_dataset fun_get_paths = lambda traj_indx: dl_scanNet.get_paths( traj_indx, frame_interv=5, split_txt=split_file, database_path_base=dataset_path) img_size = [384, 256] # trajectory index for testing # n_scenes, _, _, _, _ = fun_get_paths(0) traj_Indx = np.arange(0, n_scenes) elif dataset_name == '7scenes': # 7 scenes video # import mdataloader.dl_7scenes as dl_7scenes dataset_init = dl_7scenes.SevenScenesDataset dat_indx_step = 3 split_file = None if args.split_file == '.' else args.split_file fun_get_paths = lambda traj_indx: dl_7scenes.get_paths_1frame(\ traj_indx, database_path_base = dataset_path , split_txt = split_file, dat_indx_step = dat_indx_step) img_size = [384, 256] n_scenes, _, _, _, _ = fun_get_paths(0) traj_Indx = np.arange(0, n_scenes) elif dataset_name == 'kitti': import mdataloader.kitti as dl_kitti dataset_init = dl_kitti.KITTI_dataset fun_get_paths = lambda traj_indx: dl_kitti.get_paths( traj_indx, split_txt=split_file, mode='val') if not dataset_path == '.': fun_get_paths = lambda traj_indx: dl_kitti.get_paths( traj_indx, split_txt=split_file, mode='val', database_path_base=dataset_path) else: # use default database path fun_get_paths = lambda traj_indx: dl_kitti.get_paths( traj_indx, split_txt=split_file, mode='val') if not args.change_aspect_ratio: # we will keep the aspect ratio and do cropping img_size = [768, 256] crop_w = None else: # we will change the aspect ratio and NOT do cropping img_size = [384, 256] crop_w = None n_scenes, _, _, _, _ = fun_get_paths(0) traj_Indx = np.arange(0, n_scenes) else: raise Exception('dataset loader not implemented') fldr_path, img_paths, dmap_paths, poses, intrin_path = fun_get_paths(0) if dataset_name == 'kitti': dataset = dataset_init(True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, resize_dmap=.25, crop_w=crop_w) dataset_imgsize = dataset_init(True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, resize_dmap=1) else: dataset = dataset_init(True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, resize_dmap=.25) dataset_imgsize = dataset_init(True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, resize_dmap=1) # ================================ # print('Initnializing the KV-Net') model_KVnet = m_kvnet.KVNET(feature_dim=dnet_feature_dim, cam_intrinsics=dataset.cam_intrinsics, d_candi=d_candi, sigma_soft_max=sigma_soft_max, KVNet_feature_dim=dnet_feature_dim, d_upsample_ratio_KV_net=d_upsample, t_win_r=t_win_r, if_refined=True) model_KVnet = torch.nn.DataParallel(model_KVnet) model_KVnet.cuda() model_path_KV = args.model_path print('loading KV_net at %s' % (model_path_KV)) utils_model.load_pretrained_model(model_KVnet, model_path_KV) print('Done') for traj_idx in traj_Indx: res_fldr = '../results/%s/traj_%d' % (exp_name, traj_idx) m_misc.m_makedir(res_fldr) scene_path_info = [] print('Getting the paths for traj_%d' % (traj_idx)) fldr_path, img_seq_paths, dmap_seq_paths, poses, intrin_path = fun_get_paths( traj_idx) dataset.set_paths(img_seq_paths, dmap_seq_paths, poses) if dataset_name is 'scanNet': # For each trajector in the dataset, we will update the intrinsic matrix # dataset.get_cam_intrinsics(intrin_path) print('Done') dat_array = [dataset[idx] for idx in range(t_win_r * 2 + 1)] DMaps_meas = [] traj_length = len(dataset) print('trajectory length = %d' % (traj_length)) for frame_cnt, ref_indx in enumerate( range(t_win_r, traj_length - t_win_r - 1)): eff_iter = True valid_seq = check_datArray_pose(dat_array) # Read ref. and src. data in the local time window # ref_dat, src_dats = m_misc.split_frame_list(dat_array, t_win_r) if frame_cnt == 0: BVs_predict = None if valid_seq and eff_iter: # Get poses # src_cam_extMs = m_misc.get_entries_list_dict(src_dats, 'extM') src_cam_poses = \ [warp_homo.get_rel_extrinsicM(ref_dat['extM'], src_cam_extM_) \ for src_cam_extM_ in src_cam_extMs ] src_cam_poses = [ torch.from_numpy(pose.astype( np.float32)).cuda().unsqueeze(0) for pose in src_cam_poses ] # src_cam_poses size: N V 4 4 # src_cam_poses = torch.cat(src_cam_poses, dim=0).unsqueeze(0) src_frames = [m_misc.get_entries_list_dict(src_dats, 'img')] if frame_cnt == 0 or BVs_predict is None: # the first window for the traj. BVs_predict_in = None else: BVs_predict_in = BVs_predict print('testing on %d/%d frame in traj %d/%d ... '%\ (frame_cnt+1, traj_length - 2*t_win_r, traj_idx+1, len(traj_Indx)) ) # set trace for specific frame # BVs_measure, BVs_predict = test_KVNet.test( model_KVnet, d_candi, Ref_Dats=[ref_dat], Src_Dats=[src_dats], Cam_Intrinsics=[dataset.cam_intrinsics], t_win_r=t_win_r, Src_CamPoses=src_cam_poses, BV_predict=BVs_predict_in, R_net=True, Cam_Intrinsics_imgsize=dataset_imgsize.cam_intrinsics, ref_indx=ref_indx) # export_res.export_res_refineNet(ref_dat, BVs_measure, d_candi_dmap_ref, # res_fldr, ref_indx, # save_mat = True, output_pngs = False, output_dmap_ref=False) export_res.export_res_img(ref_dat, BVs_measure, d_candi_dmap_ref, res_fldr, frame_cnt) scene_path_info.append( [frame_cnt, dataset[ref_indx]['img_path']]) elif valid_seq is False: # if the sequence contains invalid pose estimation BVs_predict = None print('frame_cnt :%d, include invalid poses' % (frame_cnt)) elif eff_iter is False: BVs_predict = None # Update dat_array # dat_array.pop(0) dat_array.append(dataset[ref_indx + t_win_r + 1]) m_misc.save_ScenePathInfo('%s/scene_path_info.txt' % (res_fldr), scene_path_info)
def main(): import argparse print('Parsing the arguments...') parser = argparse.ArgumentParser() # exp name # parser.add_argument( '--exp_name', required=True, type=str, help='The name of the experiment. Used to naming the folders') # nepoch # parser.add_argument('--nepoch', required=True, type=int, help='# of epochs to run') # if pretrain # parser.add_argument('--pre_trained', action='store_true', default=False, help='If use the pre-trained model; (False)') # logging # parser.add_argument('--TB_add_img_interv', type=int, default=50, help='The inerval for log one training image') parser.add_argument('--pre_trained_model_path', type=str, default='.', help='The pre-trained model path for\ KV-net') # model saving # parser.add_argument( '--save_model_interv', type=int, default=5000, help='The interval of iters to save the model; default: 5000') # tensorboard # parser.add_argument( '--TB_fldr', type=str, default='runs', help='The tensorboard logging root folder; default: runs') # about training # parser.add_argument( '--RNet', action='store_true', help='if use refinement net to improve the depth resolution', default=True) parser.add_argument('--weight_var', default=.001, type=float, help='weight for the variance loss, if we use L1 loss') parser.add_argument( '--pose_noise_level', default=0, type=float, help='Noise level for pose. Used for training with pose noise') parser.add_argument('--frame_interv', default=5, type=int, help='frame interval') parser.add_argument('--LR', default=.001, type=float, help='Learning rate') parser.add_argument('--t_win', type=int, default=2, help='The radius of the temporal window; default=2') parser.add_argument('--d_min', type=float, default=0, help='The minimal depth value; default=0') parser.add_argument('--d_max', type=float, default=15, help='The maximal depth value; default=15') parser.add_argument('--ndepth', type=int, default=128, help='The # of candidate depth values; default= 128') parser.add_argument('--grad_clip', action='store_true', help='if clip the gradient') parser.add_argument('--grad_clip_max', type=float, default=2, help='the maximal norm of the gradient') parser.add_argument('--sigma_soft_max', type=float, default=500., help='sigma_soft_max, default = 500.') parser.add_argument( '--feature_dim', type=int, default=64, help='The feature dimension for the feature extractor; default=64') parser.add_argument( '--batch_size', type=int, default=0, help='The batch size for training; default=0, means batch_size=nGPU') # about dataset # parser.add_argument('--dataset', type=str, default='scanNet', help='Dataset name: {scanNet, kitti,}') parser.add_argument('--dataset_path', type=str, default='.', help='Path to the dataset') parser.add_argument( '--change_aspect_ratio', action='store_true', default=False, help= 'If we want to change the aspect ratio. This option is only useful for KITTI' ) parser.add_argument('--ngpu', type=int, default=1., help='How many GPU') # para config. # args = parser.parse_args() exp_name = args.exp_name saved_model_path = './saved_models/%s' % (exp_name) dataset_name = args.dataset if args.batch_size == 0: batch_size = torch.cuda.device_count() else: batch_size = args.batch_size n_epoch = args.nepoch TB_add_img_interv = args.TB_add_img_interv pre_trained = args.pre_trained t_win_r = args.t_win nDepth = args.ndepth d_candi = np.linspace(args.d_min, args.d_max, nDepth) d_candi_up = np.linspace(args.d_min, args.d_max, nDepth * 4) LR = args.LR sigma_soft_max = args.sigma_soft_max #10.#500. dnet_feature_dim = args.feature_dim frame_interv = args.frame_interv # should be multiple of 5 for scanNet dataset if_clip_gradient = args.grad_clip grad_clip_max = args.grad_clip_max d_candi_dmap_ref = d_candi nDepth_dmap_ref = nDepth ngpu = args.ngpu # saving model config.# m_misc.m_makedir(saved_model_path) savemodel_interv = args.save_model_interv # writer # log_dir = '%s/%s' % (args.TB_fldr, exp_name) writer = SummaryWriter(log_dir=log_dir, comment='%s' % (exp_name)) m_misc.save_args(args, '%s/tr_paras.txt' % (log_dir)) # save the training parameters # logfile = os.path.join(log_dir, 'log_' + str(time.time()) + '.txt') stdout = Logger.Logger(logfile) sys.stdout = stdout # Initialize data-loader, model and optimizer # # ===== Dataset selection ======== # dataset_path = args.dataset_path if dataset_name == 'scanNet': dataset_init = dl_scanNet.ScanNet_dataset if not dataset_path == '.': fun_get_paths = lambda traj_indx: dl_scanNet.get_paths( traj_indx, frame_interv=5, split_txt='./mdataloader/scanNet_split/scannet_train.txt', database_path_base=dataset_path) else: fun_get_paths = lambda traj_indx: dl_scanNet.get_paths( traj_indx, frame_interv=5, split_txt='./mdataloader/scanNet_split/scannet_train.txt') img_size = [384, 256] # trajectory index for training # n_scenes, _, _, _, _ = fun_get_paths(0) traj_Indx = np.arange(0, n_scenes) elif dataset_name == 'kitti': import mdataloader.kitti as dl_kitti dataset_init = dl_kitti.KITTI_dataset if not dataset_path == '.': fun_get_paths = lambda traj_indx: dl_kitti.get_paths( traj_indx, split_txt='./mdataloader/kitti_split/training.txt', mode='train', database_path_base=dataset_path) else: # use default database path fun_get_paths = lambda traj_indx: dl_kitti.get_paths( traj_indx, split_txt='./mdataloader/kitti_split/training.txt', mode='train') # img_size = [1248, 380] if not args.change_aspect_ratio: # we will keep the aspect ratio and do cropping img_size = [768, 256] crop_w = 384 else: # we will change the aspect ratio and NOT do cropping img_size = [384, 256] # img_size = [512, 256] # img_size = [624, 256] crop_w = None n_scenes, _, _, _, _ = fun_get_paths(0) traj_Indx = np.arange(0, n_scenes) else: raise Exception('dataset not implemented ') fldr_path, img_paths, dmap_paths, poses, intrin_path = fun_get_paths(0) if dataset_name == 'kitti': dataset = dataset_init(True, img_paths, dmap_paths, poses, intrin_path=intrin_path, img_size=img_size, digitize=True, d_candi=d_candi_dmap_ref, d_candi_up=d_candi_up, resize_dmap=.25, crop_w=crop_w) else: raise Exception('dataset not implemented ') # ================================ # print('Initnializing the KV-Net') model_KVnet = m_kvnet.KVNET(feature_dim=dnet_feature_dim, cam_intrinsics=dataset.cam_intrinsics, d_candi=d_candi, sigma_soft_max=sigma_soft_max, KVNet_feature_dim=dnet_feature_dim, d_upsample_ratio_KV_net=None, t_win_r=t_win_r, if_refined=args.RNet) model_KVnet = torch.nn.DataParallel(model_KVnet, device_ids=range(ngpu), dim=0) model_KVnet.cuda(0) optimizer_KV = optim.Adam(model_KVnet.parameters(), lr=LR, betas=(.9, .999)) model_path_KV = args.pre_trained_model_path if model_path_KV is not '.' and pre_trained: print('loading KV_net at %s' % (model_path_KV)) utils_model.load_pretrained_model(model_KVnet, model_path_KV, optimizer_KV) print('Done') LOSS = [] total_iter = 0 d_candi_up = d_candi for iepoch in range(n_epoch): BatchScheduler = batch_loader.Batch_Loader(batch_size=batch_size, fun_get_paths=fun_get_paths, dataset_traj=dataset, nTraj=len(traj_Indx), dataset_name=dataset_name) for batch_idx in range(len(BatchScheduler)): for frame_count, ref_indx in enumerate( range(BatchScheduler.traj_len)): local_info = BatchScheduler.local_info_full() n_valid_batch = local_info['is_valid'].sum() if n_valid_batch > 0: local_info_valid = batch_loader.get_valid_items(local_info) ref_dats_in = [] for m in range(0, len(local_info_valid['ref_dats'])): ref_dats_in.append( local_info_valid['ref_dats'][m]["left_camera"]) src_dats_in = [] for m in range(0, len(local_info_valid['src_dats'])): orig = [] for k in range(0, len(local_info_valid['src_dats'][m])): orig.append(local_info_valid['src_dats'][m][k] ["left_camera"]) src_dats_in.append(orig) cam_intrin_in = local_info_valid['left_cam_intrins'] src_cam_poses_in = torch.cat( local_info_valid['left_src_cam_poses'], dim=0) if args.pose_noise_level > 0: src_cam_poses_in = add_noise2pose( src_cam_poses_in, args.pose_noise_level) if frame_count == 0 or prev_invalid: prev_invalid = False BVs_predict_in = None print('frame_count ==0 or invalid previous frame') else: BVs_predict_in = batch_loader.get_valid_BVs( BVs_predict, local_info['is_valid']) #print("HACK") #BVs_predict_in = None BVs_measure, BVs_predict, loss, dmap_log_l, dmap_log_h = train_KVNet.train( n_valid_batch, model_KVnet, optimizer_KV, t_win_r, d_candi, Ref_Dats=ref_dats_in, Src_Dats=src_dats_in, Src_CamPoses=src_cam_poses_in, BVs_predict=BVs_predict_in, Cam_Intrinsics=cam_intrin_in, weight_var=args.weight_var, loss_type='NLL', mGPU=True) BVs_measure = BVs_measure.detach() loss_v = float(loss.data.cpu().numpy()) if n_valid_batch < BatchScheduler.batch_size: BVs_predict = batch_loader.fill_BVs_predict( BVs_predict, local_info['is_valid']) else: loss_v = LOSS[-1] prev_invalid = True # Update dat_array # if frame_count < BatchScheduler.traj_len - 1: BatchScheduler.proceed_frame() total_iter += 1 # logging # if frame_count > 0: LOSS.append(loss_v) print('video batch %d / %d, iter: %d, frame_count: %d; Epoch: %d / %d, loss = %.5f'\ %(batch_idx + 1, len(BatchScheduler), total_iter, frame_count, iepoch + 1, n_epoch, loss_v)) writer.add_scalar('data/train_error', float(loss_v), total_iter) if total_iter % savemodel_interv == 0: # if training, save the model # savefilename = saved_model_path + '/kvnet_checkpoint_iter_' + str( total_iter) + '.tar' torch.save( { 'iter': total_iter, 'frame_count': frame_count, 'ref_indx': ref_indx, 'traj_idx': batch_idx, 'state_dict': model_KVnet.state_dict(), 'optimizer': optimizer_KV.state_dict(), 'loss': loss_v }, savefilename) if total_iter % TB_add_img_interv == 0: # if training, logging # th_dmaps_log = torch.FloatTensor( dmap_log_l.astype(np.float32)) th_dmaps_log = th_dmaps_log.unsqueeze(0) th_dmaps_log = (th_dmaps_log / (d_candi_dmap_ref.max())).clamp(0, 1) th_dmaps_log = th_dmaps_log.repeat([3, 1, 1]) input_img_log = ref_dats_in[0]['img'].clone() input_img_log = (input_img_log - input_img_log.min()) / ( input_img_log.max() - input_img_log.min()) input_img_log = input_img_log.squeeze() # assuming N=1 for BVs_measure # confMap_log, _ = torch.max(BVs_measure[0, ...], dim=0) confMap_log = torch.exp(confMap_log.squeeze().cpu()) confMap_log /= confMap_log.max() confMap_log = confMap_log.repeat([3, 1, 1]) writer.add_image('%s/tr_dmaps' % (exp_name), th_dmaps_log, total_iter) writer.add_image('%s/tr_input' % (exp_name), input_img_log, total_iter) writer.add_image('%s/conf_map' % (exp_name), confMap_log, total_iter) # up-sample branch # if dmap_log_h is not -1: th_dmaps_up_log = torch.FloatTensor( dmap_log_h.astype(np.float32)) th_dmaps_up_log = th_dmaps_up_log.unsqueeze(0) th_dmaps_up_log = (th_dmaps_up_log / (d_candi_dmap_ref.max())).clamp( 0, 1) th_dmaps_up_log = th_dmaps_up_log.repeat([3, 1, 1]) writer.add_image('%s/tr_dmaps_up' % (exp_name), th_dmaps_up_log, total_iter) BatchScheduler.proceed_batch() writer.close() stdout.delink()