hofee/nbv_reconstruction
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

global_only: pipeline

Browse Source
This commit is contained in:
hofee 2024-10-29 12:04:54 +00:00
parent a21538c90a
commit f533104e4a
2 changed files with 13 additions and 40 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
configs/server/server_train_config.yaml
View File

@ -116,16 +116,16 @@ module:
feature_transform: False feature_transform: False
transformer_seq_encoder: transformer_seq_encoder:
embed_dim: 384 embed_dim: 256
num_heads: 4 num_heads: 4
ffn_dim: 256 ffn_dim: 256
num_layers: 3 num_layers: 3
output_dim: 2048 output_dim: 1024
gf_view_finder: gf_view_finder:
t_feat_dim: 128 t_feat_dim: 128
pose_feat_dim: 256 pose_feat_dim: 256
main_feat_dim: 3072 main_feat_dim: 2048
regression_head: Rx_Ry_and_T regression_head: Rx_Ry_and_T
pose_mode: rot_matrix pose_mode: rot_matrix
per_point_feature: False per_point_feature: False

47
core/pipeline.py
View File

@ -7,10 +7,10 @@ from PytorchBoot.factory.component_factory import ComponentFactory
from PytorchBoot.utils import Log from PytorchBoot.utils import Log
@stereotype.pipeline("nbv_reconstruction_global_pts_n_num_pipeline") @stereotype.pipeline("nbv_reconstruction_pipeline")
class NBVReconstructionGlobalPointsPipeline(nn.Module): class NBVReconstructionPipeline(nn.Module):
def __init__(self, config): def __init__(self, config):
super(NBVReconstructionGlobalPointsPipeline, self).__init__() super(NBVReconstructionPipeline, self).__init__()
self.config = config self.config = config
self.module_config = config["modules"] self.module_config = config["modules"]
@ -20,10 +20,6 @@ class NBVReconstructionGlobalPointsPipeline(nn.Module):
self.pose_encoder = ComponentFactory.create( self.pose_encoder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["pose_encoder"] namespace.Stereotype.MODULE, self.module_config["pose_encoder"]
) )
self.pts_num_encoder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["pts_num_encoder"]
)
self.transformer_seq_encoder = ComponentFactory.create( self.transformer_seq_encoder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["transformer_seq_encoder"] namespace.Stereotype.MODULE, self.module_config["transformer_seq_encoder"]
) )
@ -96,44 +92,21 @@ class NBVReconstructionGlobalPointsPipeline(nn.Module):
scanned_n_to_world_pose_9d_batch = data[ scanned_n_to_world_pose_9d_batch = data[
"scanned_n_to_world_pose_9d" "scanned_n_to_world_pose_9d"
] # List(B): Tensor(S x 9) ] # List(B): Tensor(S x 9)
scanned_pts_mask_batch = data[
"scanned_pts_mask"
] # Tensor(B x N)
device = next(self.parameters()).device device = next(self.parameters()).device
embedding_list_batch = [] embedding_list_batch = []
combined_scanned_pts_batch = data["combined_scanned_pts"] # Tensor(B x N x 3) combined_scanned_pts_batch = data["combined_scanned_pts"] # Tensor(B x N x 3)
global_scanned_feat, perpoint_scanned_feat_batch = self.pts_encoder.encode_points( global_scanned_feat = self.pts_encoder.encode_points(
combined_scanned_pts_batch, require_per_point_feat=True combined_scanned_pts_batch, require_per_point_feat=False
) # global_scanned_feat: Tensor(B x Dg), perpoint_scanned_feat: Tensor(B x N x Dl) ) # global_scanned_feat: Tensor(B x Dg)
for scanned_n_to_world_pose_9d, scanned_mask, perpoint_scanned_feat in zip( for scanned_n_to_world_pose_9d in scanned_n_to_world_pose_9d_batch:
scanned_n_to_world_pose_9d_batch,
scanned_pts_mask_batch,
perpoint_scanned_feat_batch,
):
scanned_target_pts_num = [] # List(S): Int
partial_feat_seq = []
seq_len = len(scanned_n_to_world_pose_9d)
for seq_idx in range(seq_len):
partial_idx_in_combined_pts = scanned_mask == seq_idx # Ndarray(V), N->V idx mask
partial_perpoint_feat = perpoint_scanned_feat[partial_idx_in_combined_pts] # Ndarray(V x Dl)
partial_feat = torch.mean(partial_perpoint_feat, dim=0) # Tensor(Dl)
partial_feat_seq.append(partial_feat)
scanned_target_pts_num.append(partial_perpoint_feat.shape[0])
scanned_target_pts_num = torch.tensor(scanned_target_pts_num, dtype=torch.float32).unsqueeze(-1).to(device) # Tensor(S x 1)
scanned_n_to_world_pose_9d = scanned_n_to_world_pose_9d.to(device) # Tensor(S x 9) scanned_n_to_world_pose_9d = scanned_n_to_world_pose_9d.to(device) # Tensor(S x 9)
pose_feat_seq = self.pose_encoder.encode_pose(scanned_n_to_world_pose_9d) # Tensor(S x Dp) pose_feat_seq = self.pose_encoder.encode_pose(scanned_n_to_world_pose_9d) # Tensor(S x Dp)
pts_num_feat_seq = self.pts_num_encoder.encode_pts_num(scanned_target_pts_num) # Tensor(S x Dn) seq_embedding = pose_feat_seq
partial_feat_seq = torch.stack(partial_feat_seq) # Tensor(S x Dl) embedding_list_batch.append(seq_embedding) # List(B): Tensor(S x (Dp))
seq_embedding = torch.cat([pose_feat_seq, pts_num_feat_seq, partial_feat_seq], dim=-1) # Tensor(S x (Dp+Dn+Dl))
embedding_list_batch.append(seq_embedding) # List(B): Tensor(S x (Dp+Dn+Dl))
seq_feat = self.transformer_seq_encoder.encode_sequence(embedding_list_batch) # Tensor(B x Ds) seq_feat = self.transformer_seq_encoder.encode_sequence(embedding_list_batch) # Tensor(B x Ds)
main_feat = torch.cat([seq_feat, global_scanned_feat], dim=-1) # Tensor(B x (Ds+Dg)) main_feat = torch.cat([seq_feat, global_scanned_feat], dim=-1) # Tensor(B x (Ds+Dg))