Skip to main content
Revision 2b019233d6851facadec8e9215cc805eef47932c authored by Changjian Chen on 20 May 2024, 01:52:04 UTC, committed by Changjian Chen on 20 May 2024, 01:52:04 UTC
update readme
1 parent 08a8fb3
Raw File Download
voc.cpython-36.pyc 
3

�L`�;�@s�dZddlmZddljZddlZddlZddlj	j
Z
ddlZddlZ
ddlZddlmZejddkrvddljjZnddljjZd$Zejed�ZGdd�de�ZGdd�de�ZGd d!�d!e
j�ZGd"d#�d#e
j�Z dS)%z�VOC Dataset Classes

Original author: Francisco Massa
https://github.com/fmassa/vision/blob/voc_dataset/torchvision/datasets/voc.py

Updated by: Ellis Brown, Max deGroot
�)�HOME�N)�logger��	aeroplane�bicycle�bird�boat�bottle�bus�car�cat�chair�cow�diningtable�dog�horse�	motorbike�person�pottedplant�sheep�sofa�train�	tvmonitorzWSL/Data/VOCdevkit/c@s"eZdZdZddd�Zdd�ZdS)	�VOCEvalAnnotationTransforma�Transforms a VOC annotation into a Tensor of bbox coords and label index
    Initilized with a dictionary lookup of classnames to indexes

    Arguments:
        class_to_ind (dict, optional): dictionary lookup of classnames -> indexes
            (default: alphabetic indexing of VOC's 20 classes)
        keep_difficult (bool, optional): keep difficult instances or not
            (default: False)
        height (int): height
        width (int): width
    NFcCs&|ptttttt����|_||_dS)N)�dict�zip�VOC_CLASSES�range�len�class_to_ind�keep_difficult)�selfr r!�r#�'/data/changjian/WSL/CSD-SSD/data/voc.py�__init__.sz#VOCEvalAnnotationTransform.__init__cCs�g}x�|jd�D]�}i}|jd�j|d<|jd�j|d<t|jd�j�|d<t|jd�j�|d<|jd�}t|jd�j�dt|jd	�j�dt|jd
�j�dt|jd�j�dg|d<|j|�qW|S)
z�
        Arguments:
            target (annotation) : the target annotation to be made usable
                will be an ET.Element
        Returns:
            a list containing lists of bounding boxes  [bbox coords, class name]
        �object�nameZposeZ	truncated�	difficult�bndbox�xminr�ymin�xmax�ymax�bbox)�findall�find�text�int�append)r"�target�width�height�objects�obj�
obj_structr.r#r#r$�__call__3s
z#VOCEvalAnnotationTransform.__call__)NF)�__name__�
__module__�__qualname__�__doc__r%r:r#r#r#r$r!s
rc@s"eZdZdZddd�Zdd�ZdS)	�VOCAnnotationTransforma�Transforms a VOC annotation into a Tensor of bbox coords and label index
    Initilized with a dictionary lookup of classnames to indexes

    Arguments:
        class_to_ind (dict, optional): dictionary lookup of classnames -> indexes
            (default: alphabetic indexing of VOC's 20 classes)
        keep_difficult (bool, optional): keep difficult instances or not
            (default: False)
        height (int): height
        width (int): width
    NFcCs&|ptttttt����|_||_dS)N)rrrrrr r!)r"r r!r#r#r$r%WszVOCAnnotationTransform.__init__cCs�g}x�|jd�D]�}t|jd�j�dk}|jr6|r6q|jd�jj�j�}|jd�}dddd	g}	g}
xNt|	�D]B\}}t|j|�j�d}
|d
dkr�|
|n|
|}
|
j|
�qnW|j	|}|
j|�||
g7}qW|S)z�
        Arguments:
            target (annotation) : the target annotation to be made usable
                will be an ET.Element
        Returns:
            a list containing lists of bounding boxes  [bbox coords, class name]
        r&r(rr'r)r*r+r,r-rr)
�iterr2r0r1r!�lower�strip�	enumerater3r )r"r4r5r6�resr8r(r'r.�ptsr)�i�ptZcur_pt�	label_idxr#r#r$r:\s"


zVOCAnnotationTransform.__call__)NF)r;r<r=r>r%r:r#r#r#r$r?Js
r?c@s�eZdZdZed gde�ddddfdd	�Zd
d�Zdd
�Zdd�Z	dd�Z
dd�Zdd�Zdd�Z
dd�Zdd�Zdd�Zdd�ZdS)!�VOCDetectionaRVOC Detection Dataset Object

    input is image, target is annotation

    Arguments:
        root (string): filepath to VOCdevkit folder.
        image_set (string): imageset to use (eg. 'train', 'val', 'test')
        transform (callable, optional): transformation to perform on the
            input image
        target_transform (callable, optional): transformation to perform on the
            target `annotation`
            (eg: take in caption string, return tensor of word indices)
        dataset_name (string, optional): which dataset to load
            (default: 'VOC2007')
    �2007�trainvalN�VOCg�?Fc	Cs�||_||_||_||_|r$t�|_t|_||_tj	ddd�|_
tj	ddd�|_|d\}	}
tj	|jd|	|
d�}|
d	kr�tj|�r�t
d
|�t�|_t|d��F}|j�jd�jd�}
x*|
D]"}|jd
�\}}|jj||f�q�WWdQRXntd��nnt
d�t�|_x\|D]T\}	}
tj	|jd|	�}x6ttj	|dd|
d��D]}|jj||j�f��qDW�qWtt|j�|�|_||_|j|�dS)Nz%s�Annotationsz%s.xml�
JPEGImagesz%s.jpgrrLz_random_list.txtrKzreading from random_list�r�
�	z#you should have randomlist.txt filezreading ids from origin data�	ImageSets�Mainz.txt)�root�	image_set�	transform�target_transformrr�classesr'�osp�join�	_annopath�_imgpath�exists�print�list�_ids�open�readrB�splitr3�
ValueErrorr2r�
supervise_num�supervise_percent�set_only_supervise)r"rT�
image_setsrVrW�dataset_namerf�only_supervise�eval�yearr'Zidlist_file�f�lists�l�a�b�rootpath�liner#r#r$r%�s>


 
 zVOCDetection.__init__cCsF||_|r,tjd�|jd|j�j�|_ntjd�|jj�|_dS)Nzusing supervised onlyz&using both supervised and unsupervised)rjr�infor`re�copy�ids)r"rjr#r#r$rg�s

zVOCDetection.set_only_supervisecCs|jS)N)rX)r"r#r#r$�
get_label_map�szVOCDetection.get_label_mapcCs|jddS)Nrr)rU)r"r#r#r$�get_set_type�szVOCDetection.get_set_typecCs|rt�|_dS)N)rrW)r"rkr#r#r$�set_eval�szVOCDetection.set_evalcCs&|j|�\}}}}}||||j|�fS)N)�	pull_item�pull_image_level_anno)r"�index�im�gt�h�w�semir#r#r$�__getitem__�szVOCDetection.__getitem__cCs
t|j�S)N)rrv)r"r#r#r$�__len__�szVOCDetection.__len__cCs|j|}tj|j|�j�}tj|j|�}|j\}}}|j	dk	rR|j	|||�}|j
dk	r�tj|�}|j
||dd�dd�f|dd�df�\}}}	|dd�dd�df}tj
|tj|	dd�f�}||jkr�tjdg�}
ntjdg�}
tj|�jddd�||||
fS)N�rrr)�axis)rrr)rv�ET�parser[�getroot�cv2�imreadr\�shaperWrV�np�array�hstack�expand_dimsre�torch�
from_numpy�permute)r"r|�img_idr4�imgr6r5�channels�boxes�labelsr�r#r#r$rz�s



0
zVOCDetection.pull_itemcCs|j|}tj|j|tj�S)aReturns the original image object at index in PIL form

        Note: not using self.__getitem__(), as any transformations passed in
        could mess up this functionality.

        Argument:
            index (int): index of img to show
        Return:
            PIL img
        )rvr�r�r\�IMREAD_COLOR)r"r|r�r#r#r$�
pull_image�s
zVOCDetection.pull_imagecCs8|j|}tj|j|�j�}|j|dd�}|d|fS)a�Returns the original annotation of image at index

        Note: not using self.__getitem__(), as any transformations passed in
        could mess up this functionality.

        Argument:
            index (int): index of img to get annotation of
        Return:
            list:  [img_id, [(label, bbox coords),...]]
                eg: ('001718', [('dog', (96, 13, 438, 332))])
        r)rvr�r�r[r�rW)r"r|r��annor~r#r#r$�	pull_annos
zVOCDetection.pull_annocCs:|j|�\}}d}tj|�}x|D]}d||d<q"W|S)N�r�����)r�r��zeros)r"r|r�r~�cat_numr��gr#r#r$r{s

z"VOCDetection.pull_image_level_annocCstj|j|��jd�S)a7Returns the original image at an index in tensor form

        Note: not using self.__getitem__(), as any transformations passed in
        could mess up this functionality.

        Argument:
            index (int): index of img to show
        Return:
            tensorized version of img, squeezed
        r)r��Tensorr��
unsqueeze_)r"r|r#r#r$�pull_tensor"szVOCDetection.pull_tensor)rJrK)r;r<r=r>�VOC_ROOTr?r%rgrwrxryr�r�rzr�r�r{r�r#r#r#r$rI{s"2		rIc@sVeZdZdZdgde�dfdd�Zdd	�Zd
d�Zdd
�Zdd�Z	dd�Z
dd�ZdS)�VOC0712DetectionaRVOC Detection Dataset Object

    input is image, target is annotation

    Arguments:
        root (string): filepath to VOCdevkit folder.
        image_set (string): imageset to use (eg. 'train', 'val', 'test')
        transform (callable, optional): transformation to perform on the
            input image
        target_transform (callable, optional): transformation to perform on the
            target `annotation`
            (eg: take in caption string, return tensor of word indices)
        dataset_name (string, optional): which dataset to load
            (default: 'VOC2007')
    rJrKNZVOC0712c
Cs�||_||_||_||_||_tjddd�|_tjddd�|_t	�|_
xX|D]P\}}tj|jd|�}x4ttj|dd|d	��D]}	|j
j||	j
�f�q�WqLWdS)
Nz%srMz%s.xmlrNz%s.jpgrLrRrSz.txt)rTrUrVrWr'rYrZr[r\r_rvrar3rB)
r"rTrhrVrWrirlr'rrrsr#r#r$r%CszVOC0712Detection.__init__cCs|j|�\}}}}||fS)N)rz)r"r|r}r~rr�r#r#r$r�TszVOC0712Detection.__getitem__cCs
t|j�S)N)rrv)r"r#r#r$r�YszVOC0712Detection.__len__c
Cs�|j|}tj|j|�j�}tj|j|�}|j\}}}|j	dk	rR|j	|||�}|j
dk	r�tj|�}|j
||dd�dd�f|dd�df�\}}}	|dd�dd�df}tj
|tj|	dd�f�}tj|�jddd�|||fS)Nr�rrr)r�)rrr)rvr�r�r[r�r�r�r\r�rWrVr�r�r�r�r�r�r�)
r"r|r�r4r�r6r5r�r�r�r#r#r$rz\s



0zVOC0712Detection.pull_itemcCs|j|}tj|j|tj�S)aReturns the original image object at index in PIL form

        Note: not using self.__getitem__(), as any transformations passed in
        could mess up this functionality.

        Argument:
            index (int): index of img to show
        Return:
            PIL img
        )rvr�r�r\r�)r"r|r�r#r#r$r�ps
zVOC0712Detection.pull_imagecCs8|j|}tj|j|�j�}|j|dd�}|d|fS)a�Returns the original annotation of image at index

        Note: not using self.__getitem__(), as any transformations passed in
        could mess up this functionality.

        Argument:
            index (int): index of img to get annotation of
        Return:
            list:  [img_id, [(label, bbox coords),...]]
                eg: ('001718', [('dog', (96, 13, 438, 332))])
        r)rvr�r�r[r�rW)r"r|r�r�r~r#r#r$r�~s
zVOC0712Detection.pull_annocCstj|j|��jd�S)a7Returns the original image at an index in tensor form

        Note: not using self.__getitem__(), as any transformations passed in
        could mess up this functionality.

        Argument:
            index (int): index of img to show
        Return:
            tensorized version of img, squeezed
        r)r�r�r�r�)r"r|r#r#r$r��szVOC0712Detection.pull_tensor)rJrK)r;r<r=r>r?r%r�r�rzr�r�r�r#r#r#r$r�0sr�)rrrr	r
rrr
rrrrrrrrrrrr)!r>�configr�os.path�pathrY�sysr��torch.utils.data�utils�datar��numpyr��random�utils.loggerr�version_infoZxml.etree.cElementTree�etreeZcElementTreer��xml.etree.ElementTree�ElementTreerrZr�r&rr?�DatasetrIr�r#r#r#r$�<module>s.
)16