![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*rE3bf23O8Ulf4szSw0-EMg.jpeg\")
Visualizing data to carry out a detailed EDA, especially for image data, is critical. In computer vision<\/strong> datasets, if we can view and compare the images across different views with their relevant metadata and transformations within a single and well-designed UI, we are one step ahead in solving a CV task. We<\/p>\n\n\n\n In this article, we will explore the following:<\/p>\n\n\n\n There are various ways we can construct a Kangas DataGrid. Check out this tutorial on Constructing and visualizing Kangas DataGrid on Kangas UI<\/a>.<\/p>\n\n\n\n We can store images on a DataGrid using the Kangas Here we will use an image of a cat and load it using the Kangas provides a The image data passed into the See other formats<\/a>.<\/p>\n\n\n\n Image data as NumPy matrix:<\/p>\n\n\n\n Image data as TensorFlow tensor:<\/p>\n\n\n\n We can load the image to a DataGrid and visualize it with the UI.<\/p>\n\n\n\n When we create images, we can specify more data describing various attributes of the images we are working with. By default, Kangas logs out the We can visualize this on the UI. Click on the image, and a pop window will appear. Locate the \u201cMetadata\u201d section and toggle the dropdown.<\/em><\/p>\n\n\n\n Note <\/strong>that the The Image annotations like bounding boxes and masks are included as metadata.<\/p>\n\n\n\n In object detection tasks, bounding boxes are essential as they help us determine the spatial location of a particular object.<\/p>\n\n\n\n The Kangas Image masking lets us focus on the parts of the image we are most interested in. For instance, if we build a computer vision model to recognize faces, we can mask the rest of the image and focus on the actual faces.<\/p>\n\n\n\n While we can do image masking in various ways, for example, using OpenCV, Kangas has the With this method, a mask can be:<\/p>\n\n\n\n We can visualize and apply built-in functions from Kangas UI on hugging face datasets<\/a> with Kangas. Hugging Face contains a collection of datasets we can use to build various models.<\/p>\n\n\n\n In this section, we will see how to create DataGrid from hugging face datasets and visualize them on the Kangas UI. We will grab the license plate object detection dataset<\/a>.<\/p>\n\n\n\n First, import the datasets library to enable us to grab the dataset:<\/p>\n\n\n\n Next, import the following:<\/p>\n\n\n\n For the simplicity of this article, we will only split 20 examples from the training set of the dataset.<\/p>\n\n\n\n We have 5 features, and we have split 20 examples.\n
Creating DataGrid with images.<\/h1>\n\n\n\n
Image()<\/code> class or as Pillow images(PIL).<\/p>\n\n\n\nLoading images with Image() class<\/h2>\n\n\n\n
Image()<\/code> class.<\/p>\n\n\n\nimage = kg.Image('\/content\/cats-dogs\/cat1.jpeg'<\/span>)\nprint<\/span>('TYPE OF IMAGE: '<\/span>, type<\/span>(image))\nprint<\/span>()\nimage.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:418\/1*elu5dcdt5I2n5boUvm8Iow.png\")
Load image as Pillow (PIL) image with Kangas<\/h2>\n\n\n\n
to_pil()<\/code> method that converts the image into a Pillow image.<\/p>\n\n\n\nimage = kg.Image('\/content\/cats-dogs\/cat1.jpeg'<\/span>).to_pil()\nprint<\/span>('TYPE OF IMAGE: '<\/span>, type<\/span>(image))\nprint<\/span>()\nimage.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:461\/1*b8s5NyWclmO76X5YpVJdXA.png\")
Image()<\/code> class can take a variety of formats. Let\u2019s try to load an image as a NumPy matrix and a TensorFlow Tensor. View more formats on the Creating an Image Asset<\/p>\n\n\n\n# Image<\/span>\nimage = kg.Image('\/content\/cats-dogs\/cat1.jpeg'<\/span>).to_pil()<\/span><\/pre>\n\n\n\nfrom<\/span> numpy import<\/span> asarray\n\nnumpy_img_data = asarray(image)\nprint<\/span>(f\"Type of data: {type<\/span>(numpy_img_data)}<\/span> data: of Shape: {numpy_img_data.shape}<\/span>\"<\/span>)\n\nprint<\/span>('TYPE OF IMAGE: '<\/span>, type<\/span>(image))\nprint<\/span>()\nimage.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:533\/1*bkgsPfoN2wCcnfr1OP79SQ.png\")
<\/figure>\n\n\n\nimport<\/span> tensorflow as<\/span> tf\n\ntensor_img_data = tf.convert_to_tensor(image)\nprint<\/span>(f\"Type of data: {type<\/span>(tensor_img_data)}<\/span> data: of Shape: {tensor_img_data.shape}<\/span>\"<\/span>)\n\nprint<\/span>('TYPE OF IMAGE: '<\/span>, type<\/span>(image))\nprint<\/span>()\nimage.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:687\/1*7g1RnD2-QGkgfr9pTNOIpg.png\")
<\/figure>\n\n\n\nAdding images to a DataGrid<\/h2>\n\n\n\n
image = kg.Image('\/content\/cats-dogs\/cat1.jpeg'<\/span>, name='cat.jpeg'<\/span>) # provide name to be used as filename in metadata<\/span>\ndg = kg.DataGrid(name='Cat'<\/span>) # provide name to be used as the datagrid name<\/span>\ndg.extend([{'image'<\/span>: image}])\n\n# Fire up the UI<\/span>\ndg.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*oa4yxs40cvKYxnd2qHWwjQ.png\")
import<\/span> glob\n\n#instantiate DataGrid<\/span>\ndg = DataGrid(name=\"Images\"<\/span>, columns=[\"image_file_path\"<\/span>, \"image\"<\/span>, 'height'<\/span>, 'width'<\/span>, \"label\"<\/span>])\n\n# append data to DataGrid<\/span>\nfor<\/span> file in<\/span> glob.iglob(\"\/*\/*\/*.jpeg\"<\/span>):\n file_name = os.path.basename(file) # file_name to metadata<\/span>\n image = Image.open<\/span>(file)\n image_w = image.width\n image_h = image.height\n label = os.path.splitext(file_name)[0<\/span>][:3<\/span>]\n dg.append([file, kg.Image(file, name=file_name), image_w, image_h, label])<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*433aiR1E04WgKHFTH4p2tA.png\")
Adding Image metadata<\/h1>\n\n\n\n
name(as <\/code>filename) and theassetId<\/code>.<\/p>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*JlgJjoGIX8f7mBpCUPi5CA.png\")
extension <\/code>value is derived from the filename <\/code>value. If the filename(name argument)<\/code>is not provided, the value would be the file path.<\/p>\n\n\n\nImage()<\/a><\/code> class<\/a> provides more optional arguments that we can provide to give more information about the image.<\/p>\n\n\n\nAdding bounding boxes to image with Kangas<\/h2>\n\n\n\n
\n
![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*WOkBhqQjEX7UMNejqXxagQ.png\")
add_bounding_boxes()<\/a><\/code> method can receive either of the above representations. We can pass the following to the method:<\/p>\n\n\n\n\n
img2_aug = kg.Image('\/content\/cat_dog.jpg'<\/span>)\n\nbox1 = [(40<\/span>, 65<\/span>), (210<\/span>, 280<\/span>)]\nbox2 = [(222<\/span>, 23<\/span>),(448<\/span>, 281<\/span>)]\nimgs_bbox = img2_aug.add_bounding_boxes(\"Cat\"<\/span>, box1, score=0.89<\/span>, layer_name=\"Truth\"<\/span>, id<\/span>='cat01'<\/span>)\nimgs_bbox = img2_aug.add_bounding_boxes(\"Dog\"<\/span>, box2, score=0.89<\/span>, layer_name=\"Truth\"<\/span>, id<\/span>='dog01'<\/span>)\n\nimg_dg = kg.DataGrid(name='cat_dog'<\/span>)\nimg_dg.extend([[imgs_bbox]])\n\nimg_dg.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*H68e5aGm6V8MTVGv5J8JAg.gif\")
Image Masking with Kangas<\/h2>\n\n\n\n
add_mask()<\/code> method we can use when creating an image.<\/p>\n\n\n\n\n
np.array<\/code> with int values.<\/li>\n\n\n\nkangas.Image image<\/code>.<\/li>\n<\/ul>\n\n\n\nimage = kg.Image('\/content\/cat_dog.jpg'<\/span>)\nmask = PIL.Image.open<\/span>('\/content\/mask_truth.png'<\/span>)\nimage.add_mask({1<\/span>: 'Cat_vs_dog'<\/span>}, mask, layer_name='Truth'<\/span>)\ndg = kg.DataGrid(name=\"model1\"<\/span>)\ndg.append([image])\ndg.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*rd6hOfoAnlakVT1GPQS_Iw.png\")
Loading and visualizing image data from hugging face.<\/h1>\n\n\n\n
![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*rCB1CZKXlhK3Mn9xzrh7Kg.png\")
%pip install datasets<\/span><\/pre>\n\n\n\nimport<\/span> datasets\nfrom<\/span> datasets import<\/span> load_datasets, list_datasets\n<\/span><\/pre>\n\n\n\nlicense_dataset = load_dataset('keremberke\/license-plate-object-detection'<\/span>, 'full'<\/span>, split='train[:20]'<\/span>)\nlicense_dataset<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:547\/1*vvqQdi_jiaQ6uBPBoYb6Iw.png\")
<\/figure>\n\n\n\n
\nLet\u2019s load the datasets into a Kangas DataGrid and visualize it in the UI:<\/p>\n\n\n\ndg_obj_detection = kg.DataGrid(licence_plate_dataset, name='Licence_plate_obj_detection'<\/span>)\n\ndg_obj_detection.show()<\/span><\/pre>\n\n\n\n![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*DjEXVvjaoAdtND9C8P4-Ag.png\")
![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:439\/1*FCZK9J_rXI6kuhH2eG5YwA.png\")
![\"\"\/](\"https:\/\/miro.medium.com\/v2\/resize:fit:700\/1*r64nG6KxGz5msoBbodaXkA.gif\")