|
| 1 | +# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. |
| 2 | +# |
| 3 | +# Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | +# you may not use this file except in compliance with the License. |
| 5 | +# You may obtain a copy of the License at |
| 6 | +# |
| 7 | +# http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | +# |
| 9 | +# Unless required by applicable law or agreed to in writing, software |
| 10 | +# distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | +# See the License for the specific language governing permissions and |
| 13 | +# limitations under the License. |
| 14 | + |
| 15 | +import os |
| 16 | + |
| 17 | +from .dataset import Dataset |
| 18 | +from paddleseg.cvlibs import manager |
| 19 | +from paddleseg.transforms import Compose |
| 20 | + |
| 21 | + |
| 22 | +@manager.DATASETS.add_component |
| 23 | +class PPHumanSeg14K(Dataset): |
| 24 | + """ |
| 25 | + This is the PP-HumanSeg14K Dataset. |
| 26 | +
|
| 27 | + This dataset was introduced in the work: |
| 28 | + Chu, Lutao, et al. "PP-HumanSeg: Connectivity-Aware Portrait Segmentation with a Large-Scale Teleconferencing Video Dataset." Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2022. |
| 29 | +
|
| 30 | + This dataset is divided into training set, validation set and test set. The training set includes 8770 pictures, the validation set includes 2431 pictures, and the test set includes 2482 pictures. |
| 31 | +
|
| 32 | + Args: |
| 33 | + dataset_root (str, optional): The dataset directory. Default: None. |
| 34 | + transforms (list, optional): Transforms for image. Default: None. |
| 35 | + mode (str, optional): Which part of dataset to use. It is one of ('train', 'val'). Default: 'train'. |
| 36 | + edge (bool, optional): Whether to compute edge while training. Default: False. |
| 37 | + """ |
| 38 | + NUM_CLASSES = 2 |
| 39 | + |
| 40 | + def __init__(self, |
| 41 | + dataset_root=None, |
| 42 | + transforms=None, |
| 43 | + mode='train', |
| 44 | + edge=False): |
| 45 | + self.dataset_root = dataset_root |
| 46 | + self.transforms = Compose(transforms) |
| 47 | + mode = mode.lower() |
| 48 | + self.mode = mode |
| 49 | + self.file_list = list() |
| 50 | + self.num_classes = self.NUM_CLASSES |
| 51 | + self.ignore_index = 255 |
| 52 | + self.edge = edge |
| 53 | + |
| 54 | + if mode not in ['train', 'val', 'test']: |
| 55 | + raise ValueError( |
| 56 | + "`mode` should be 'train', 'val' or 'test', but got {}.".format( |
| 57 | + mode)) |
| 58 | + |
| 59 | + if self.transforms is None: |
| 60 | + raise ValueError("`transforms` is necessary, but it is None.") |
| 61 | + |
| 62 | + if mode == 'train': |
| 63 | + file_path = os.path.join(self.dataset_root, 'train.txt') |
| 64 | + elif mode == 'val': |
| 65 | + file_path = os.path.join(self.dataset_root, 'val.txt') |
| 66 | + else: |
| 67 | + file_path = os.path.join(self.dataset_root, 'test.txt') |
| 68 | + |
| 69 | + with open(file_path, 'r') as f: |
| 70 | + for line in f: |
| 71 | + items = line.strip().split(' ') |
| 72 | + if len(items) != 2: |
| 73 | + if mode == 'train' or mode == 'val': |
| 74 | + raise Exception( |
| 75 | + "File list format incorrect! It should be" |
| 76 | + " image_name label_name\\n") |
| 77 | + image_path = os.path.join(self.dataset_root, items[0]) |
| 78 | + grt_path = None |
| 79 | + else: |
| 80 | + image_path = os.path.join(self.dataset_root, items[0]) |
| 81 | + grt_path = os.path.join(self.dataset_root, items[1]) |
| 82 | + self.file_list.append([image_path, grt_path]) |
0 commit comments