File size: 6,356 Bytes
9affda1 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 | import pandas as pd
import os
import json
import glob
# 1. 基础配置
repro_csv = "outputs/phase5a/task_vapre_splatatlas_repro.csv"
output_submission = "outputs/phase5a/appendix_B_scene_inventory_submission_ready.csv"
output_audit = "outputs/phase5a/appendix_B_scene_inventory_audit.csv"
if not os.path.exists(repro_csv):
print(f"Error: 找不到基础文件 {repro_csv}")
exit(1)
df_repro = pd.read_csv(repro_csv)
valid_scenes = sorted(df_repro['scene'].unique().tolist())
# 2. 知识库映射字典
dataset_map = {}
resolution_map = {}
type_map = {}
display_name_map = {}
# Mip-NeRF 360 Indoor
for s in ["bonsai", "counter", "kitchen", "room"]:
dataset_map[s] = "Mip-NeRF 360"
resolution_map[s] = "2"
type_map[s] = "real-world indoor"
display_name_map[s] = s.capitalize()
# Mip-NeRF 360 Outdoor
for s in ["bicycle", "flowers", "garden", "stump", "treehill"]:
dataset_map[s] = "Mip-NeRF 360"
resolution_map[s] = "4"
type_map[s] = "real-world outdoor"
display_name_map[s] = s.capitalize()
# Tanks & Temples
tnt_indoor = ["auditorium", "ballroom", "courtroom", "museum", "palace", "temple"]
tnt_outdoor = ["barn", "caterpillar", "truck", "playground", "lighthouse", "train"]
for s in tnt_indoor + tnt_outdoor:
dataset_map[s] = "Tanks&Temples"
resolution_map[s] = "2"
type_map[s] = "real-world indoor" if s in tnt_indoor else "real-world outdoor"
display_name_map[s] = s.capitalize()
# NeRF-Synthetic
for s in ["chair", "drums", "ficus", "hotdog", "lego", "materials", "mic", "ship"]:
dataset_map[s] = "NeRF-Synthetic"
resolution_map[s] = "1"
type_map[s] = "synthetic object"
display_name_map[s] = s.capitalize()
# Deep Blending
for s in ["playroom", "drjohnson"]:
dataset_map[s] = "Deep Blending"
resolution_map[s] = "1"
type_map[s] = "real-world indoor"
display_name_map[s] = "Playroom" if s == "playroom" else "DrJohnson"
# 3. 提取引擎
inventory = []
for s in valid_scenes:
ds = dataset_map.get(s, "UNKNOWN")
res = resolution_map.get(s, "NEEDS_MANUAL_CHECK")
stype = type_map.get(s, "UNKNOWN")
dname = display_name_map.get(s, s)
train_views = "NEEDS_MANUAL_CHECK"
test_views = "NEEDS_MANUAL_CHECK"
evidence = "inferred_from_dataset_convention"
notes = ""
# 尝试从 NeRF-Synthetic 目录读取 view 数量
if ds == "NeRF-Synthetic":
# 考虑到可能的大小写目录名
for scene_dir in [s, s.capitalize()]:
train_json = f"/root/autodl-tmp/dataset/nerf_synthetic_off/nerf_synthetic/{scene_dir}/transforms_train.json"
test_json = f"/root/autodl-tmp/dataset/nerf_synthetic_off/nerf_synthetic/{scene_dir}/transforms_test.json"
if os.path.exists(train_json) and os.path.exists(test_json):
try:
with open(train_json, 'r') as f:
train_data = json.load(f)
train_views = str(len(train_data['frames']))
with open(test_json, 'r') as f:
test_data = json.load(f)
test_views = str(len(test_data['frames']))
evidence = f"parsed_from_transforms_json"
break
except Exception:
pass
# 若未能解析，则统一归为需手工确认
inventory.append({
"SceneKey": s,
"DisplayName": dname,
"Dataset": ds,
"TrainViews": train_views,
"TestViews": test_views,
"Resolution": res,
"SceneType": stype,
"SourceEvidence": evidence,
"Notes": notes
})
df_inv = pd.DataFrame(inventory)
# 4. 导出表格
os.makedirs("outputs/phase5a", exist_ok=True)
# 投稿版
sub_cols = ["SceneKey", "DisplayName", "Dataset", "TrainViews", "TestViews", "Resolution", "SceneType"]
df_inv[sub_cols].to_csv(output_submission, index=False)
# 审计版
audit_cols = ["SceneKey", "DisplayName", "Dataset", "TrainViews", "TestViews", "Resolution", "SceneType", "SourceEvidence", "Notes"]
df_inv[audit_cols].to_csv(output_audit, index=False)
print(f"========================================================")
print(f"Submission 级表格已保存至: {output_submission}")
print(f"Audit 级审计表已保存至: {output_audit}")
print(f"========================================================\n")
# 5. Markdown 预览
print("=== Appendix B: Scene Inventory (Preview) ===")
print(df_inv[sub_cols].to_markdown(index=False))
# 6. Sanity Checks
print("\n========================================================")
print("SANITY CHECKS")
print("========================================================")
print(f"1. Scene 总数是否正好 31: {len(df_inv) == 31} ({len(df_inv)})")
c2 = all(s in df_repro['scene'].unique().tolist() for s in df_inv['SceneKey'])
print(f"2. SceneKey 是否和 repro CSV 完全一致: {c2}")
allowed_datasets = ["Tanks&Temples", "Deep Blending", "Mip-NeRF 360", "NeRF-Synthetic"]
c3 = df_inv['Dataset'].isin(allowed_datasets).all()
print(f"3. Dataset 是否全部属于四类之一: {c3}")
print(f"4. 四个 Dataset 的 scene count 分布:")
for k, v in df_inv['Dataset'].value_counts().items():
print(f" - {k}: {v}")
c5 = df_inv['Resolution'].notnull().all() and not (df_inv['Resolution'] == "NEEDS_MANUAL_CHECK").any()
print(f"5. Resolution 是否全部非空且无 NEEDS_MANUAL_CHECK: {c5}")
missing_train = (df_inv['TrainViews'] == "NEEDS_MANUAL_CHECK").sum()
missing_test = (df_inv['TestViews'] == "NEEDS_MANUAL_CHECK").sum()
print(f"6. TrainViews 缺失数量 (NEEDS_MANUAL_CHECK): {missing_train}")
print(f"7. TestViews 缺失数量 (NEEDS_MANUAL_CHECK): {missing_test}")
inferred_count = (df_inv['SourceEvidence'] == 'inferred_from_dataset_convention').sum()
print(f"8. 字段为 inferred_from_dataset_convention 的数量: {inferred_count}")
# 风险检测
capitalization_risks = ["lego", "drjohnson", "playroom"]
print("\n9. 大小写风险检查 (SceneKey vs DisplayName):")
for r in capitalization_risks:
row = df_inv[df_inv['SceneKey'] == r]
if not row.empty:
print(f" - {r} -> {row.iloc[0]['DisplayName']}")
print("\n10. 需要手动补全 View 数量的场景:")
manual_views = df_inv[df_inv['TrainViews'] == 'NEEDS_MANUAL_CHECK']['SceneKey'].tolist()
print(f" - {', '.join(manual_views)}")
|