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

update transformer_seq_encoder's config

Browse Source
This commit is contained in:
hofee 2024-10-06 13:53:32 +08:00
parent fa69f9f879
commit bfc8ba0f4b
2 changed files with 11 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
configs/server/server_train_config.yaml
View File

@ -90,7 +90,7 @@ pipeline:
nbv_reconstruction_global_pts_pipeline: nbv_reconstruction_global_pts_pipeline:
modules: modules:
pts_encoder: pointnet_encoder pts_encoder: pointnet_encoder
pose_seq_encoder: transformer_pose_seq_encoder pose_seq_encoder: transformer_seq_encoder
pose_encoder: pose_encoder pose_encoder: pose_encoder
view_finder: gf_view_finder view_finder: gf_view_finder
eps: 1e-5 eps: 1e-5
@ -107,20 +107,12 @@ module:
feature_transform: False feature_transform: False
transformer_seq_encoder: transformer_seq_encoder:
pts_embed_dim: 1024 embed_dim: 1344
pose_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: 2048
transformer_pose_seq_encoder:
pose_embed_dim: 256
num_heads: 4
ffn_dim: 256
num_layers: 3
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

15
core/global_pts_n_num_pipeline.py
View File

@ -12,21 +12,24 @@ class NBVReconstructionGlobalPointsPipeline(nn.Module):
super(NBVReconstructionGlobalPointsPipeline, self).__init__() super(NBVReconstructionGlobalPointsPipeline, self).__init__()
self.config = config self.config = config
self.module_config = config["modules"] self.module_config = config["modules"]
self.pts_encoder = ComponentFactory.create( self.pts_encoder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["pts_encoder"] namespace.Stereotype.MODULE, self.module_config["pts_encoder"]
) )
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.pose_n_num_seq_encoder = ComponentFactory.create( self.pts_num_encoder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["pose_n_num_seq_encoder"] namespace.Stereotype.MODULE, self.module_config["pts_num_encoder"]
)
self.transformer_seq_encoder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["transformer_seq_encoder"]
) )
self.view_finder = ComponentFactory.create( self.view_finder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["view_finder"] namespace.Stereotype.MODULE, self.module_config["view_finder"]
) )
self.pts_num_encoder = ComponentFactory.create(
namespace.Stereotype.MODULE, self.module_config["pts_num_encoder"]
)
self.eps = float(self.config["eps"]) self.eps = float(self.config["eps"])
self.enable_global_scanned_feat = self.config["global_scanned_feat"] self.enable_global_scanned_feat = self.config["global_scanned_feat"]
@ -128,7 +131,7 @@ class NBVReconstructionGlobalPointsPipeline(nn.Module):
seq_embedding = torch.cat([pose_feat_seq, pts_num_feat_seq, partial_feat_seq], dim=-1) # Tensor(S x (Dp+Dn+Dl)) 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)) embedding_list_batch.append(seq_embedding) # List(B): Tensor(S x (Dp+Dn+Dl))
seq_feat = self.pose_n_num_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))