from detectron2.utils.visualizer import Visualizer # image and prediction outputs img = cv2.imread("image.jpg") outputs = ... # initializing visualizer object v = Visualizer(img[:, :, ::-1], metadata=None, scale=1) # drawing prediction results on image v = v.draw_instance_predictions(outputs["instances"].to("cpu")) cv2.imshow("Detection Results", v.get_image()[:, :, ::-1])
from detectron2.utils.visualizer import Visualizer, ColorMode # image and prediction outputs img1 = cv2.imread("image1.jpg") img2 = cv2.imread("image2.jpg") outputs1 = ... outputs2 = ... # initializing visualizer object for each image v1 = Visualizer(img1[:, :, ::-1], metadata=None, scale=1, instance_mode=ColorMode.IMAGE_BW) v2 = Visualizer(img2[:, :, ::-1], metadata=None, scale=1, instance_mode=ColorMode.IMAGE_BW) # drawing prediction results on images v1 = v1.draw_instance_predictions(outputs1["instances"].to("cpu")) v2 = v2.draw_instance_predictions(outputs2["instances"].to("cpu")) # displaying images side by side concatenated_image = np.concatenate((v1.get_image()[:, :, ::-1], v2.get_image()[:, :, ::-1]), axis=1) cv2.imshow("Detection Results", concatenated_image)In this example, we read two input images and obtain the output detection results using a model. We then initialize two Visualizer objects, one for each image, and draw the predicted instances onto the images. Finally, we concatenate the modified images side by side using the `np.concatenate` function and display the concatenated image using the `cv2.imshow` function. The detectron2.utils.visualizer Visualizer class is part of the detectron2 package library, which is an open-source framework for building object detection models. The library is built on top of PyTorch and provides various utilities for data loading, model training, and evaluation of object detection models.