Question

我正在从事计算机视觉工作，我正在尝试将Microsoft Coco数据集转换为png图像，以便我可以直接在Caffe上使用它们。

我修改了它们的API的一些功能，以便我有正确的分割颜色（每个类id在rgb中有（id，id，id）颜色，而background是（0,0,0））。这是修改后的功能：

def showAnns(self, anns):
    """
    Display the specified annotations.
    :param anns (array of object): annotations to display
    :return: None
    """
    if len(anns) == 0:
        return 0
    if 'segmentation' in anns[0]:
        datasetType = 'instances'
    elif 'caption' in anns[0]:
        datasetType = 'captions'
    if datasetType == 'instances':
        ax = plt.gca()
        polygons = []
        color = []
        # sort annotations from biggest to smallest to avoid occlusions
        anns.sort(key=lambda x: x['area'], reverse=True)

        for ann in anns:

            pixelvalue = ann['category_id']/255.0
            c = [pixelvalue, pixelvalue, pixelvalue]
            if type(ann['segmentation']) == list:
                # polygon
                for seg in ann['segmentation']:
                    poly = np.array(seg).reshape((len(seg)/2, 2))
                    polygons.append(Polygon(poly, True,alpha=1))
                    color.append(c)
            else:
                # mask
                t = self.imgs[ann['image_id']]
                if type(ann['segmentation']['counts']) == list:
                    rle = mask.frPyObjects([ann['segmentation']], t['height'], t['width'])
                else:
                    rle = [ann['segmentation']]
                m = mask.decode(rle)
                img = np.ones( (m.shape[0], m.shape[1], 3) )
                color_mask = c
                for i in range(3):
                    img[:,:,i] = color_mask[i]
                ax.imshow(np.dstack( (img, m*0.5) ))

        p = PatchCollection(polygons, facecolors=color, edgecolors=(1,1,1,1), linewidths=1, alpha=1)
        ax.add_collection(p)
    elif datasetType == 'captions':
        for ann in anns:
            print ann['caption']

此代码在我的python脚本中调用：

from pycocotools.coco import COCO
import numpy as np
import skimage.io as io
import matplotlib.pyplot as plt

dataDir='..'
dataType='train2014'
annFile='%s/annotations/instance`enter code here`s_%s.json'%(dataDir,dataType)

# initialize COCO api for instance annotations
coco=COCO(annFile)

#display COCO categories and supercategories
cats = coco.loadCats(coco.getCatIds())

# get all images
catIds = coco.getCatIds();
imgIds = coco.getImgIds()
imgs = coco.loadImgs(imgIds)
imgs.sort(key=lambda x: x['id'])

for img in imgs:

    # load and display image
    I = io.imread('%s/images/%s/%s'%(dataDir,dataType,img['file_name']))
    # 0 all values
    I[:] = 0

    axes = plt.gca()
    axes.set_frame_on(False)
    axes.set_xticks([]); axes.set_yticks([])

    plt.axis('off')
    plt.imshow(I)

    # load and display instance annotations
    annIds = coco.getAnnIds(imgIds=img['id'], catIds=catIds, iscrowd=None)
    anns = coco.loadAnns(annIds)
    coco.showAnns(anns)
    path='vocstyle_cocoimages/'+str(img['file_name'])[:-4]+'.png'
    plt.savefig(path, bbox_inches='tight', pad_inches=0)
    plt.clf()

所以这里是我的问题：当我这样做时，我获得了一个与原始版本不同的保存png。我希望具有完全相同的尺寸，因为它对于评估很重要：任何错位的像素都会降低深度学习阶段的准确性。

我还没有找到一个正确的方法来保存＆＃34;图像＆＃34;该图的一部分被剥去了图中的任何其他东西。我们说我有一个640 * 480的图像输入，我想要与输出相同的尺寸。

感谢您阅读

Answer 1

使用计数器，并将其附加到路径字符串：

COUNT = 0

for img in imgs:

    # load and display image
    I = io.imread('%s/images/%s/%s'%(dataDir,dataType,img['file_name']))
    # 0 all values for black background
    I[:] = 0
    plt.imshow(I)
    # load and display instance annotations
    annIds = coco.getAnnIds(imgIds=img['id'], catIds=catIds, iscrowd=None)
    anns = coco.loadAnns(annIds)
    coco.showAnns(anns)
    print img['file_name']
    #path='vocstyle_cocoimages/'+str(img['file_name'])[:-4]+'.png'
    #your new line
    path='vocstyle_cocoimages/'+str(img['file_name'])[:-4]+ '_' + str(COUNT) + '.png'
    plt.savefig(path, bbox_inches='tight', pad_inches=0)
    #increment counter
    COUNT = COUNT + 1

仅在matplotlib的循环中将图像保存为png，具有相同的尺寸

1 个答案: