-
Notifications
You must be signed in to change notification settings - Fork 41
/
more_commands.sh
138 lines (117 loc) · 4.19 KB
/
more_commands.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
#!/usr/bin/env bash
# In this file, we provide commands to get the quantitative results presented in the MVSplat paper.
# Commands are provided by following the order of Tables appearing in the paper.
# --------------- Default Final Models ---------------
# Table 1: re10k
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/re10k.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true
# Table 1: acid
python -m src.main +experiment=acid \
checkpointing.load=checkpoints/acid.ckpt \
mode=test \
dataset/view_sampler=evaluation \
dataset.view_sampler.index_path=assets/evaluation_index_acid.json \
test.compute_scores=true
# generate video
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/re10k.ckpt \
mode=test \
dataset/view_sampler=evaluation \
dataset.view_sampler.index_path=assets/evaluation_index_re10k_video.json \
test.save_video=true \
test.save_image=false \
test.compute_scores=false
# --------------- Cross-Dataset Generalization ---------------
# Table 2: RealEstate10K -> ACID
python -m src.main +experiment=acid \
checkpointing.load=checkpoints/re10k.ckpt \
mode=test \
dataset/view_sampler=evaluation \
dataset.view_sampler.index_path=assets/evaluation_index_acid.json \
test.compute_scores=true
# Table 2: RealEstate10K -> DTU (2 context views)
python -m src.main +experiment=dtu \
checkpointing.load=checkpoints/re10k.ckpt \
mode=test \
dataset/view_sampler=evaluation \
dataset.view_sampler.index_path=assets/evaluation_index_dtu_nctx2.json \
test.compute_scores=true
# RealEstate10K -> DTU (3 context views)
python -m src.main +experiment=dtu \
checkpointing.load=checkpoints/re10k.ckpt \
mode=test \
dataset/view_sampler=evaluation \
dataset.view_sampler.index_path=assets/evaluation_index_dtu_nctx3.json \
dataset.view_sampler.num_context_views=3 \
wandb.name=dtu/views3 \
test.compute_scores=true
# --------------- Ablation Models ---------------
# Table 3: base
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_worefine.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_base \
model.encoder.wo_depth_refine=true
# Table 3: w/o cost volume
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_worefine_wocv.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_wocv \
model.encoder.wo_depth_refine=true \
model.encoder.wo_cost_volume=true
# Table 3: w/o cross-view attention
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_worefine_wobbcrossattn_best.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_wo_backbone_cross_attn \
model.encoder.wo_depth_refine=true \
model.encoder.wo_backbone_cross_attn=true
# Table 3: w/o U-Net
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_worefine_wounet.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_wo_unet \
model.encoder.wo_depth_refine=true \
model.encoder.wo_cost_volume_refine=true
# Table B: w/ Epipolar Transformer
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_worefine_wepitrans.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_w_epipolar_trans \
model.encoder.wo_depth_refine=true \
model.encoder.use_epipolar_trans=true
# Table C: 3 Gaussians per pixel
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_gpp3.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_gpp3 \
model.encoder.gaussians_per_pixel=3
# Table D: w/ random init (300K)
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_wopretrained.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_wo_pretrained
# Table D: w/ random init (450K)
python -m src.main +experiment=re10k \
checkpointing.load=checkpoints/ablations/re10k_wopretrained_450k.ckpt \
mode=test \
dataset/view_sampler=evaluation \
test.compute_scores=true \
wandb.name=abl/re10k_wo_pretrained_450k