{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"***\n\n","metadata":{}},{"cell_type":"markdown","source":"# 0. Introduction\n\nImages and videos are very important in various fields of science and technology. For example, in cancer diagnosis, security, entertainment and autonomous cars. By entering these fields, artificial intelligence (AI) has brought such a change in people's lives that today's life is like a science-fiction movie compared to ten or twenty years ago. Despite all these developments, every year the growth of knowledge in this field is more than before. Therefore, in this essay, the progress of artificial intelligence in technology and science in the field of images and videos from 2022 to 2023 will be discussed.","metadata":{}},{"cell_type":"markdown","source":"# 1. What is computer vision?\n\nComputer vision, a subfield of computer science and artificial inteligence, tries to identify, analyze, compare or even generate the objects, people, animals and various concepts based on data recorded by sensors and cameras using computational tools. Creating a meaningful connection and understanding of the raw data of the images or videos, depends on the model of artificial intelligence having a correct understanding of the concepts related to the segments of that data. This understanding usually happens by presenting a large amount of similar data as an artificial intelligence training stage, and in some models, the association of those images with natural language or other types of data is also established.\n\nTo create meaningful concepts and understanding of an AI model, it is necessary to go through some steps so that the AI model can work accurately. In the initial stage of this process, images or videos are recorded with tools such as sensors or cameras, which is called image acquisition. Then, in the pre-processing stage, the stored data should be cleaned and slightly edited so that its quality increases and the AI model can understand it as easily as possible. For example, photos are cropped, their noise is reduced, and their scale and color elements are changed. Then its output is given to the AI model, which today is mostly an artificial neural network, so that it can be trained according to that processed input data. At this stage, it may be necessary to give a label to each image to identify its components by the AI model and establish a connection between the image and its meaning (for example, its corresponding word in a natural language), which can be time-consuming. In the stage of training or testing that AI model, it must first perform an operation called feature extraction in order to be able to recognize the edges of image components and to recognize lines, general shapes and textures. Then, using the features obtained according to the previous step, a task is performed to prepare the final output, which is called patter recognition.\n\nCommon computer vision tasks include image classification, object tracking or detection, and facial recognition. For each of these tasks, special architectures have been built for the structure of artificial neural networks and different types of approaches and processes have been proposed to reduce its error and increase its speed and accuracy. Finally, these outputs are used in industries such as self-driving cars, healthcare, manufacturing, design and digital fabrication. The main purpose of this essay is to review and briefly introduce the architectures and approaches of the last two years (2022 and 2023) and its latest uses in these industries.\n","metadata":{}},{"cell_type":"markdown","source":"# 2. History\nThe history of images and videos in AI goes back decades. In the 1960s, researchers began to develop algorithms for pattern recognition techniques. In the next two decades, this research led to rule-based systems. During this time, the aim of most researches has been to identify simple shapes in images.\n\nIn the 1990s, with the advent of machine learning algorithms, research took a new direction. For example, techniques such as Support Vector Machines (SVMs) and Neural Networks were able to automatically analyze images. Towards the end of this period, Convolutional Neural Networks (CNNs) ushered in a new wave of capabilities in this field.\n\nIn the 2000s, the situation changed completely with the advent of more powerful and faster computers, as well as the ability to store large amounts of information. Several algorithms were developed in this decade that performed well in the field of object recognition. Also, the analysis of simple images expanded to the analysis of videos and actions in video frames.\n\nThe astonishing development in these few decades continued more steeply until in the 2010s, with the capabilities of Deep Learning in artificial intelligence, the state of innovation resembled the science fiction projections of 50 years ago. Architectures like AlexNet and GoogLeNet were able to perform very powerfully. It was in this decade that with generative models such as GANs and VAEs, content and data generation by artificial intelligence became possible.\n","metadata":{}},{"cell_type":"markdown","source":"# 3. Recent Developments\nIn 2022, some AI models such as Midjourney, Dall_E 2, and Stable Diffusion managed to stun everyone from artists, engineers, and content creators. Not much time has passed since the presence of AI generated pictures of cute cats, jokes with celebrities, and the artistic use of architects, painters, and clothing designers in social media or even academic papers, when we reached new advancements in 2023. It can be said that in these years, generative artificial intelligence models were more popular than anything else. Although areas such as object recognition and surveillance in autonomous cars and security systems are very important and have taken a large share of the use of Computer Vision in the industry. In the following, this essay is going to examine the use of Computer Vision in fields such as industry, art, security, healthcare, and entertainment.","metadata":{}},{"cell_type":"markdown","source":"## Applications\n\n### Transportation\nTransportation, as a very important topic in industry and urban planning problems, has been able to use the capabilities of computer vision in recent years, because many of its issues can be evaluated using visual data (the noise of cars, level of pollution or air temperature considered as exceptions that cannot be evaluated well using computer vision). Some of the uses of computer vision in transportation include:\n* autonomous vehicles\n* Pedestrian detection and tracking\n* parking occupancy detection\n* traffic flow analysis\n* infrastructure condition monitoring\n\n### Healthcare\nMuch of the data in the health field is of visual type, so one of the biggest uses of computer vision has happened in this field. Examining MRI and CT images can help identify the disease and prevent its worsening. Also, images of the body, such as examining skin spots and visible diseases, allow each person to have a quick examination of their body. These uses can be important from different aspects. For example, the speed of diagnosis and its accuracy increases, and financially, it can help the vulnerable portion of the society and help them with their initial medical examinations, as well as meet the immediate needs of patients who do not have easy access to a doctor. The different types of these uses are listed below:\n* X-Ray, CT and MRI analysis\n* cancer detection\n* blood loss measurement\n* pathology\n* movement analysis\n\n### Editing\nIn 2023, Adobe announced that it has created a feature for editing videos in the popular Premiere Pro software, where the designer can tell the artificial intelligence his request by writing text so that the video can be edited automatically.\n\nAnother project that can be useful for content creators is Nvidia's artificial intelligence model that can create eye contact in videos. With this AI model, you can look at the text in your hand and even though you are not actually looking at the camera, the video is made as if you are staring directly at the viewer.","metadata":{}},{"cell_type":"markdown","source":"***","metadata":{}}]}