search

PRBonn / SHINE_mapping

🌟 SHINE-Mapping: Large-Scale 3D Mapping Using Sparse Hierarchical Implicit Neural Representations (ICRA 2023)
MIT License
448 stars 32 forks source link

what shuold I do to make this program run all frame on mai_city sequence 00? #15

Closed Zhangjyhhh closed 1 year ago

Zhangjyhhh commented 1 year ago

setting: name: "maicity_incre_reg" output_root: "/home/hh/experiments/" pc_path: "/home/hh/data/mai_city/ply/sequences/01/velodyne" pose_path: /home/hh/data/mai_city/ply/sequences/01/poses.txt" calib_path: "/home/hh/data/mai_city/ply/sequences/01/calib.txt" load_model: False # load the pretrained decoder model (optional) model_path: "./pretrained/geo_decoder_8dim.pth" first_frame_ref: False begin_frame: 0 end_frame: 699 every_frame: 1 # 1 means does not skip device: "cuda" gpu_id: "0" process: min_range_m: 1.5 pc_radius_m: 20.0 # distance filter for each frame rand_downsample: False # use random or voxel downsampling vox_down_m: 0.03 rand_down_r: 0.2 sampler: surface_sample_range_m: 0.5 surface_sample_n: 3 free_sample_begin_ratio: 0.3 free_sample_end_dist_m: 0.8 free_sample_n: 3 octree: leaf_vox_size: 0.5 tree_level_world: 12 tree_level_feat: 3 feature_dim: 8 poly_int_on: True octree_from_surface_samples: True decoder: mlp_level: 2 mlp_hidden_dim: 32 freeze_after_frame: 20 loss: ray_loss: False main_loss_type: sdf_bce # select from sdf_bce (our proposed), sdf_l1, sdf_l2, dr, dr_neus sigma_sigmoid_m: 0.05 loss_weight_on: False behind_dropoff_on: False ekional_loss_on: False weight_e: 0.1 continual: continual_learning_reg: True # using incremental mapping with regularization lambda_forget: 1e6 # the larger this value, the model would be less likely to forget window_replay_on: False # replay within the sliding window window_radius_m: 0 optimizer: iters: 50 # iterations per frame batch_size: 4096 learning_rate: 0.01 weight_decay: 0 # l2 regularization eval: wandb_vis_on: False # log to wandb or not o3d_vis_on: True # visualize the mapping or not vis_freq_iters: 0 save_freq_iters: 0 # save the model and octree every x iterations mesh_freq_frame: 5 # reconstruct the mesh every x frames mc_res_m: 0.2 # reconstruction marching cubes resolution mc_with_octree: False # querying sdf in the map bbx mc_vis_level: 1 save_map: False # save the sdf map or not thanks for your excellent work. I just want to know that why I change the parameter "end frame" from 100 to 699 in files config/maicity/maicity_incre_reg, I get the same result when "end frame=100". I just want to run all frame in mai_city sequence 00.May need to modify another parameter too,could you tell me

YuePanEdward commented 1 year ago

Thanks for your interest in our work. MaiCity sequence 01 only has 100 frames. You may need to change pc_path and pose_path to xxx/sequences/00/xxx to use MaiCity sequence 00, which contains 700 frames.

Zhangjyhhh commented 1 year ago

Thanks for your interest in our work. MaiCity sequence 01 only has 100 frames. You may need to change pc_path and pose_path to xxx/sequences/00/xxx to use MaiCity sequence 00, which contains 700 frames.

thanks for your reply