|
SCENE_DIR: examples/apple/ |
|
|
|
|
|
|
|
|
|
_target_: vggt.models.vggt.VGGT |
|
|
|
num_register_tokens: 4 |
|
ffn_layer: "mlp" |
|
qk_norm: False |
|
patch_size: 14 |
|
init_values: 0.01 |
|
|
|
AGGREGATOR: |
|
_target_: vggt.models.aggregator.Aggregator |
|
patch_embed_by_conv: False |
|
image_size: 518 |
|
use_checkpoint: True |
|
use_reentrant: False |
|
decoder_load_dino: False |
|
backbone_qk_norm: False |
|
aa_block_kwargs: |
|
dim: 1024 |
|
num_heads: 16 |
|
mlp_ratio: 4 |
|
qkv_bias: True |
|
proj_bias: True |
|
ffn_bias: True |
|
drop: 0.0 |
|
attn_drop: 0.0 |
|
init_values: 0.01 |
|
drop_path: 0.0 |
|
fused_attn: True |
|
qk_norm: True |
|
rope_freq: 100 |
|
|
|
|
|
CameraHead: |
|
_target_: vggt.heads.camera_head.CameraHead |
|
pose_encoding_type: "absT_quaR_FoV" |
|
new_trunk: True |
|
trunk_depth: 4 |
|
|
|
qk_norm: True |
|
init_values: 0.01 |
|
act_dict: |
|
trans_act: "linear" |
|
quat_act: "linear" |
|
fl_act: "linear" |
|
loss_kwargs: |
|
loss_type: "l1" |
|
gamma: 0.6 |
|
|
|
|
|
PointHead: |
|
_target_: vggt.heads.dpt_head.DPTHead |
|
|
|
dim_in: 2048 |
|
shallow_conv: False |
|
normalize_act: "inv_log" |
|
pos_embed: True |
|
loss_kwargs: |
|
gradient_loss: "normal" |
|
|
|
normalize_pred: False |
|
valid_range: 0.98 |
|
gamma: 1.0 |
|
camera_centric_reg: -1.0 |
|
all_mean: True |
|
|
|
DepthHead: null |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
MatchHead: null |
|
TrackHead: null |
|
|
|
|
|
|
|
hydra: |
|
output_subdir: NULL |
|
run: |
|
dir: . |
|
|