File size: 6,147 Bytes
271f3e5 | 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 | #!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""Generate OBJ-DINO and TVA manifest JSONs from an HOI eval dataset."""
import argparse
import json
import os
import os.path as osp
_EVAL_ROOT = osp.abspath(osp.join(osp.dirname(__file__), ".."))
_DEFAULT_DATASET = osp.join(
_EVAL_ROOT, "benchmarks/datasets/hoi_crosstest_20251107.json"
)
_DEFAULT_OBJ_TABLE = osp.join(
_EVAL_ROOT, "benchmarks/datasets/prompt_translated3.json"
)
CAPTION_OBJ_NAME_FALLBACK = {
"selfcollect250509_img_1-1": "notebook",
"selfcollect250509_img_1-2": "notebook",
"selfcollect250509_img_4": "electric guitar",
"selfcollect250509_img_5-1": "tablet",
"selfcollect250509_img_5-2": "cup",
"selfcollect250509_img_7": "container",
"selfcollect250509_img_8": "smartphone",
"selfcollect250509_img_13": "laptop",
"selfcollect250509_img_14": "tablet",
"selfcollect250509_img_15": "camera",
"selfcollect250509_img_17": "paper cup",
"selfcollect250509_img_18": "glass of water",
"selfcollect250509_img_23": "shoes",
"selfcollect250509_img_25": "smartphone",
"selfcollect250509_img_27": "tablet",
"selfcollect250509_img_29-1": "handbag",
"selfcollect250509_img_29-2": "handbag",
"selfcollect250509_img_31-1": "sandal",
"selfcollect250509_img_31-2": "loafer",
"selfcollect250509_img_32": "jar",
"selfcollect250509_img_33": "tablet",
"selfcollect250509_img_38": "smartphone",
"selfcollect250509_img_42": "cup",
"selfcollect250509_img_48": "plastic bag",
"selfcollect250509_img_51-1": "compact mirror",
"selfcollect250509_img_51-2": "compact mirror",
"selfcollect250509_img_53": "acoustic guitar",
"selfcollect250509_img_54": "acoustic guitar",
"selfcollect250509_img_57": "smartphone",
"selfcollect250509_img_62": "electric guitar",
"selfcollect250509_img_67": "acoustic guitar",
"selfcollect250509_img_70": "water bottle",
"selfcollect250509_img_72": "water bottle",
"selfcollect250509_img_75": "laptop",
"selfcollect250509_img_76": "sneakers",
"selfcollect250509_img_81": "smartphone",
"selfcollect250509_img_85": "compact mirror",
"selfcollect250509_img_92": "smartphone",
"selfcollect250509_img_101": "smartphone",
"selfcollect250509_img_103": "tablet",
}
def convert_obj_table(obj_data):
return {item["id"]: item for item in obj_data}
def extract_obj_name_from_path(data):
for key in ("ref_obj_image_path", "ref_image_path"):
path = data.get(key) or ""
for marker in ("_##object##_", "_##human##_"):
if marker in path:
suffix = path.split(marker)[-1]
if suffix.endswith(".png"):
obj_raw = suffix[:-4].strip()
if obj_raw:
return obj_raw.replace("_", " ")
return None
def build_obj_dino_list(hoi_list, obj_dict, skip_missing_obj=False, fallback_obj_from_path=False):
out = []
for data in hoi_list:
video_id = data["image_id"]
parts = video_id.split("=")
obj_id = parts[1] if len(parts) >= 2 else video_id
if obj_id in obj_dict:
obj_name = obj_dict[obj_id]["obj_en"]
else:
if fallback_obj_from_path:
obj_name = extract_obj_name_from_path(data)
if obj_name is None:
obj_name = CAPTION_OBJ_NAME_FALLBACK.get(video_id)
if obj_name is None:
if skip_missing_obj:
continue
raise KeyError(
f"obj_id '{obj_id}' not in obj table (video_id={video_id})"
)
elif skip_missing_obj:
continue
else:
raise KeyError(f"obj_id '{obj_id}' not in obj table (video_id={video_id})")
out.append({
"video_id": data["image_id"],
"mp4_img": data["ref_image_path"],
"obj_img": data["ref_obj_image_path"],
"obj_name": obj_name,
})
return out
def build_tva_list(hoi_list):
return [
{"video_id": item["image_id"], "prompt": item["caption"]}
for item in hoi_list
]
def main():
parser = argparse.ArgumentParser(
description="Generate OBJ-DINO and TVA manifests from an HOI eval JSON"
)
parser.add_argument("--input", type=str, default=_DEFAULT_DATASET)
parser.add_argument("--obj_table", type=str, default=_DEFAULT_OBJ_TABLE)
parser.add_argument("--obj_dino_out", type=str, default=None)
parser.add_argument("--tva_out", type=str, default=None)
parser.add_argument("--skip_missing_obj", action="store_true")
parser.add_argument("--fallback_obj_from_path", action="store_true")
args = parser.parse_args()
if args.obj_dino_out is None:
args.obj_dino_out = osp.join(
_EVAL_ROOT, "benchmarks/manifests/obj_dino_crosstest.json"
)
if args.tva_out is None:
args.tva_out = osp.join(
_EVAL_ROOT, "benchmarks/manifests/tva_crosstest.json"
)
with open(args.input, "r", encoding="utf-8") as f:
hoi_list = json.load(f)
print(f"[prepare] loaded {len(hoi_list)} samples from {args.input}")
with open(args.obj_table, "r", encoding="utf-8") as f:
obj_dict = convert_obj_table(json.load(f))
obj_dino_list = build_obj_dino_list(
hoi_list,
obj_dict,
skip_missing_obj=args.skip_missing_obj,
fallback_obj_from_path=args.fallback_obj_from_path,
)
os.makedirs(osp.dirname(args.obj_dino_out), exist_ok=True)
with open(args.obj_dino_out, "w", encoding="utf-8") as f:
json.dump(obj_dino_list, f, ensure_ascii=False, indent=4)
print(f"[prepare] OBJ_DINO -> {args.obj_dino_out} ({len(obj_dino_list)} items)")
tva_list = build_tva_list(hoi_list)
os.makedirs(osp.dirname(args.tva_out), exist_ok=True)
with open(args.tva_out, "w", encoding="utf-8") as f:
json.dump(tva_list, f, ensure_ascii=False, indent=4)
print(f"[prepare] TVA -> {args.tva_out} ({len(tva_list)} items)")
if __name__ == "__main__":
main()
|