Spaces:
Runtime error
Runtime error
File size: 4,978 Bytes
a57c6eb |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 |
import os
from typing import Dict, List
from ...utils.path_util import get_file_name_ext
from ...utils.util import load_dct_from_file
from .vision_object import Object
from .vision_frame import Frame, FrameSeq
def face_meta_2_tme_meta(src: dict) -> dict:
"""人脸中的元信息格式转换
Args:
src (dict): 人脸中的元信息
Returns:
dict: 转换后的元信息
"""
dst = {}
dst["media_name"] = src["video_name"]
dst["mediaid"] = src["video_name"]
dst["signature"] = src["video_file_hash_code"]
dst["fps"] = src["fps"]
dst["duration"] = src["duration"]
dst["frame_num"] = src["frame_num"]
dst["height"] = src["height"]
dst["width"] = src["width"]
return dst
def face_obj_2_tme_obj(src: dict) -> dict:
"""人脸信息转换为 Object中的元信息
Args:
src (dict): 人脸框相关信息
Returns:
dict: 转换后的人脸信息
"""
obj = {}
obj["category"] = "face"
obj["bbox"] = src["bbox"]
obj["kps"] = src["kps"]
obj["det_score"] = src["det_score"]
obj["gender"] = src["gender"]
obj["age"] = src["age"]
obj["trackid"] = src["roleid"]
return obj
def face_clips_2_tme_clips(src: list) -> list:
"""人脸信息转换为Clip
Args:
src (list): 人脸中 Clip 的多帧检测信息
Returns:
list: Clip 中的 frames信息
"""
dst = []
for idx, frame_perception in enumerate(src):
frame_dst = {}
frame_dst["frame_idx"] = frame_perception["frame_idx"]
objs = []
if frame_perception["faces"] is not None:
for face in frame_perception["faces"]:
obj = face_obj_2_tme_obj(face)
objs.append(obj)
frame_dst["objs"] = objs
dst.append(frame_dst)
return dst
def face2TMEType(src: dict) -> dict:
"""人脸检测的信息转换成 视频剪辑中的格式
Args:
src (dict): 人脸检测信息
Returns:
dict: 转换后的字典格式
"""
meta_info = face_meta_2_tme_meta(
{
k: v
for k, v in src.items()
if k
not in [
"face_detections",
"single_frame_transiton_score",
"all_frame_transiton_score",
"clips",
]
}
)
clips = face_clips_2_tme_clips(src["face_detections"])
video_info = {"meta_info": meta_info, "sub_meta_info": [], "clips": clips}
return video_info
def load_multi_face(
path_lst: str,
) -> dict:
"""读取多个人脸检测结果文件，转化成VideoInfo对应的字典格式。
Args:
path_lst (str or [str]): 人脸检测结果文件
Returns:
dict: VideoInfo对应的字典格式, key是 文件名
"""
if not isinstance(path_lst, list):
path_lst = [path_lst]
face_info_dct = {}
for path in path_lst:
filename, ext = get_file_name_ext(os.path.basename(path))
face_info = load_dct_from_file(path)
face_info = face2TMEType(face_info)
face_info_dct[filename] = face_info
return face_info_dct
def face_roles2frames(src: dict, **kwargs: dict) -> List[Frame]:
"""将roles字典转换为Frame
Args:
src (dict): {
roleid: {
"bbox": {
"frame_idx": [
[x1, y1, x2, y2]
]
}
"names": str,
}
}
kwargs (dict): 便于其他需要的参数也传到Frame中去
Returns:
List[Frame]: _description_
"""
frames = {}
for roleid, faces_info in src.items():
if "name" not in faces_info or faces_info["name"] == "":
name = "unknown"
else:
name = faces_info["name"]
if "bbox" in faces_info:
frames_bbox = faces_info["bbox"]
for frameid, bbox in frames_bbox.items():
frameid = int(frameid)
if frameid not in frames:
frames[frameid] = {"objs": [], "frame_idx": frameid}
obj = {
"name": name,
"bbox": bbox[0],
"category": "person",
"obj_id": int(roleid),
}
obj = Object(**obj)
frames[frameid]["objs"].append(obj)
frame_obj_list = []
for frameid in sorted(frames.keys()):
frame_args = frames[frameid]
frame_args.update(**kwargs)
frame = Frame(**frame_args)
frame_obj_list.append(frame)
return frame_obj_list
def clipseq_face_roles2frames(clips_roles: List[Dict], **kwargs: dict) -> FrameSeq:
frame_seq = []
for roles in clips_roles:
frames = face_roles2frames(roles)
frame_seq.extend(frames)
frame_seq = sorted(frame_seq, key=lambda f: f.frame_idx)
return FrameSeq(frame_seq, **kwargs)
|