File size: 4,276 Bytes
4c4ff57
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
from collections import defaultdict
from dataclasses import dataclass, field
from enum import Enum
from typing import Dict, List, Optional, Tuple

import pandas as pd

IMAGES_EXT: Tuple = (".jpeg", ".jpg", ".png", ".webp", ".bmp", ".gif")
VIDEO_EXT: Tuple = (".mp4", ".avi", ".mov", ".mkv", ".webm")


@dataclass
class PictureInfo:
    image_path: str
    age: Optional[str]  # age or age range(start;end format) or "-1"
    gender: Optional[str]  # "M" of "F" or "-1"
    bbox: List[int] = field(default_factory=lambda: [-1, -1, -1, -1])  # face bbox: xyxy
    person_bbox: List[int] = field(default_factory=lambda: [-1, -1, -1, -1])  # person bbox: xyxy

    @property
    def has_person_bbox(self) -> bool:
        return any(coord != -1 for coord in self.person_bbox)

    @property
    def has_face_bbox(self) -> bool:
        return any(coord != -1 for coord in self.bbox)

    def has_gt(self, only_age: bool = False) -> bool:
        if only_age:
            return self.age != "-1"
        else:
            return not (self.age == "-1" and self.gender == "-1")

    def clear_person_bbox(self):
        self.person_bbox = [-1, -1, -1, -1]

    def clear_face_bbox(self):
        self.bbox = [-1, -1, -1, -1]


class AnnotType(Enum):
    ORIGINAL = "original"
    PERSONS = "persons"
    NONE = "none"

    @classmethod
    def _missing_(cls, value):
        print(f"WARN: Unknown annotation type {value}.")
        return AnnotType.NONE


def get_all_files(path: str, extensions: Tuple = IMAGES_EXT):
    files_all = []
    for root, subFolders, files in os.walk(path):
        for name in files:
            # linux tricks with .directory that still is file
            if "directory" not in name and sum([ext.lower() in name.lower() for ext in extensions]) > 0:
                files_all.append(os.path.join(root, name))
    return files_all


class InputType(Enum):
    Image = 0
    Video = 1
    VideoStream = 2


def get_input_type(input_path: str) -> InputType:
    if os.path.isdir(input_path):
        print("Input is a folder, only images will be processed")
        return InputType.Image
    elif os.path.isfile(input_path):
        if input_path.endswith(VIDEO_EXT):
            return InputType.Video
        if input_path.endswith(IMAGES_EXT):
            return InputType.Image
        else:
            raise ValueError(
                f"Unknown or unsupported input file format {input_path}, \
                             supported video formats: {VIDEO_EXT}, \
                             supported image formats: {IMAGES_EXT}"
            )
    elif input_path.startswith("http") and not input_path.endswith(IMAGES_EXT):
        return InputType.VideoStream
    else:
        raise ValueError(f"Unknown input {input_path}")


def read_csv_annotation_file(annotation_file: str, images_dir: str, ignore_without_gt=False):
    bboxes_per_image: Dict[str, List[PictureInfo]] = defaultdict(list)

    df = pd.read_csv(annotation_file, sep=",")

    annot_type = AnnotType("persons") if "person_x0" in df.columns else AnnotType("original")
    print(f"Reading {annotation_file} (type: {annot_type})...")

    missing_images = 0
    for index, row in df.iterrows():
        img_path = os.path.join(images_dir, row["img_name"])
        if not os.path.exists(img_path):
            missing_images += 1
            continue

        face_x1, face_y1, face_x2, face_y2 = row["face_x0"], row["face_y0"], row["face_x1"], row["face_y1"]
        age, gender = str(row["age"]), str(row["gender"])

        if ignore_without_gt and (age == "-1" or gender == "-1"):
            continue

        if annot_type == AnnotType.PERSONS:
            p_x1, p_y1, p_x2, p_y2 = row["person_x0"], row["person_y0"], row["person_x1"], row["person_y1"]
            person_bbox = list(map(int, [p_x1, p_y1, p_x2, p_y2]))
        else:
            person_bbox = [-1, -1, -1, -1]

        bbox = list(map(int, [face_x1, face_y1, face_x2, face_y2]))
        pic_info = PictureInfo(img_path, age, gender, bbox, person_bbox)
        assert isinstance(pic_info.person_bbox, list)

        bboxes_per_image[img_path].append(pic_info)

    if missing_images > 0:
        print(f"WARNING: Missing images: {missing_images}/{len(df)}")
    return bboxes_per_image, annot_type