{"id":51,"date":"2024-09-28T14:43:23","date_gmt":"2024-09-28T06:43:23","guid":{"rendered":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/?page_id=51"},"modified":"2024-09-29T00:19:16","modified_gmt":"2024-09-28T16:19:16","slug":"%e6%99%ba%e6%85%a7ai%e6%a8%a1%e5%9e%8b","status":"publish","type":"page","link":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/%e6%99%ba%e6%85%a7ai%e6%a8%a1%e5%9e%8b\/","title":{"rendered":"\u667a\u6167AI\u6a21\u578b"},"content":{"rendered":"\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t
\u4eba\u5de5\u667a\u6167(artificial intelligence, AI)\u6307\u80fd\u6a21\u64ec\u6216\u8907\u88fd\u4eba\u985e\u667a\u6167\u7684\u4e00\u7a2e\u6280\u8853\uff0c\u958b\u767c\u9019\u7a2e\u6280\u8853\u4f7f\u96fb\u8166\u80fd\u5920\u50cf\u4eba\u4e00\u6a23\u601d\u8003\u3001\u5b78\u7fd2\u3001\u89e3\u6c7a\u554f\u984c\u548c\u505a\u51fa\u6c7a\u7b56\uff0c\u4eba\u985e\u667a\u6167\u7684\u884c\u70ba\u8af8\u5982\u6a21\u4eff\u3001\u5b78\u7fd2\u3001\u601d\u8003\u3001\u5224\u65b7\u4ee5\u53ca\u555f\u767c\u7b49\uff0c\u56e0\u6b64\u4eba\u5de5\u667a\u6167\u662f\u4e00\u7a2e\u7406\u5ff5\u3001\u76ee\u7684\u6216\u662f\u6846\u67b6\uff0c\u53ef\u4ee5\u56ca\u62ec\u8af8\u591a\u6982\u5ff5\u8207\u65b9\u6cd5\u3002\u4eba\u5de5\u667a\u6167\u53ef\u4ee5\u5206\u70ba\u591a\u500b\u5b50\u9818\u57df\uff0c\u6bcf\u500b\u5b50\u9818\u57df\u90fd\u5c08\u6ce8\u65bc\u4e0d\u540c\u7684\u554f\u984c\u548c\u6280\u8853\uff0c\u5728\u4eba\u5de5\u667a\u6167\u7684\u57fa\u790e\u4e0b\u8a95\u751f\u4e86\u8a31\u591a\u8af8\u5982\u8cc7\u6599\u63a2\u52d8(data mining)\u3001\u6a5f\u5668\u5b78\u7fd2(machine learning)\u4ee5\u53ca\u6df1\u5ea6\u5b78\u7fd2(deep learning)\u7b49\u6982\u5ff5\u53ca\u65b9\u6cd5\uff0c\u9019\u4e9b\u4eba\u5de5\u667a\u6167\u65b9\u6cd5\u5f7c\u6b64\u76f8\u95dc\u806f\uff0c\u76ee\u6a19\u7686\u70ba\u8b93\u6a5f\u5668\u64c1\u6709\u90e8\u5206\u65bc\u4eba\u985e\u667a\u6167\u76f8\u4f3c\u7684\u80fd\u529b\u3002<\/br><\/br>\nArtificial Intelligence (AI) refers to a technology that simulates or replicates human intelligence, enabling computers to think, learn, solve problems, and make decisions like humans. AI involves behaviors such as imitation, learning, reasoning, judgment, and inspiration. Therefore, AI can be viewed as a concept, a goal, or a framework that encompasses various ideas and methods. AI can be divided into several subfields, each focusing on different problems and techniques. Concepts and methods such as data mining, machine learning, and deep learning have emerged based on the foundation of AI. These AI methods are interrelated and aim to provide machines with human-like cognitive abilities.<\/br><\/br>\n\n\u8cc7\u6599\u63a2\u52d8\u65b9\u6cd5\u900f\u904e\u5404\u7a2e\u4e0d\u540c\u65b9\u6cd5\u4f86\u6316\u6398\u539f\u672c\u5927\u91cf\u4e14\u8907\u96dc\u7684\u8cc7\u6599\uff0c\u76ee\u7684\u5728\u7d66\u5b9a\u7684\u8cc7\u6599\u4e2d\u6316\u6398\u77e5\u8b58\u3001\u7279\u5fb5\u6216\u95dc\u4fc2\u3002\u8cc7\u6599\u63a2\u52d8\u65b9\u6cd5\u5305\u542b\u4e86\u8a31\u591a\u7a2e\u6a5f\u5668\u5b78\u7fd2\u65b9\u6cd5\u8207\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\uff0c\u53ea\u8981\u662f\u57fa\u65bc\u5927\u91cf\u8cc7\u6599\u4e2d\u5206\u6790\u7684\u65b9\u6cd5\u90fd\u80fd\u7a31\u4f5c\u70ba\u8cc7\u6599\u63a2\u52d8\u65b9\u6cd5\u3002\u6a5f\u5668\u5b78\u7fd2\u662f\u5728\u5df2\u8655\u7406\u904e\u7684\u8cc7\u6599\u4e2d\u900f\u904e\u5b78\u7fd2\u5df2\u77e5\u7684\u7279\u5fb5\u4f86\u5c0d\u65b0\u4e8b\u7269\u9032\u884c\u5206\u6790\u6216\u9810\u6e2c\uff0c\u6a5f\u5668\u5b78\u7fd2\u5728\u9032\u884c\u5b78\u7fd2\u524d\u5fc5\u9808\u7d66\u5b9a\u4e00\u7d44\u5df2\u7531\u4eba\u985e\u77e5\u8b58\u8403\u53d6\u51fa\u7279\u5fb5\u4e26\u6574\u7406\u5f8c\u7684\u8cc7\u6599\u3002<\/br><\/br>\n\nData mining techniques utilize various methods to extract knowledge, features, or relationships from complex and massive datasets. It encompasses a range of machine learning and deep learning methods. Any approach that analyzes large datasets can be referred to as data mining. Machine learning involves analyzing or predicting new data based on known features extracted from pre-processed data. Before learning, machine learning requires a set of data that has been labeled and organized with features extracted from human knowledge.<\/br><\/br>\n\n\u800c\u6df1\u5ea6\u5b78\u7fd2\u57fa\u65bc\u795e\u7d93\u7db2\u8def(neural network)\u7406\u8ad6\uff0c\u4e0d\u540c\u65bc\u6a5f\u5668\u5b78\u7fd2\u65b9\u6cd5\uff0c\u6df1\u5ea6\u5b78\u7fd2\u5927\u5e45\u6e1b\u5c11\u9700\u5148\u7531\u4eba\u985e\u77e5\u8b58\u4f86\u505a\u7684\u77e5\u8b58\u8403\u53d6\uff0c\u800c\u662f\u4ee5\u5927\u91cf\u7684\u795e\u7d93\u5143\u5efa\u69cb\u76f8\u7576\u8907\u96dc\u7684\u795e\u7d93\u7db2\u8def\uff0c\u4e26\u5f9e\u5927\u91cf\u8cc7\u6599\u4e2d\u8b93\u591a\u5c64\u7d50\u69cb\u7684\u795e\u7d93\u7db2\u8def\u5b78\u7fd2\u8cc7\u6599\u4e2d\u96b1\u542b\u7684\u7279\u5fb5\u70ba\u4f55\uff0c\u5c0d\u8cc7\u6599\u9032\u884c\u5b78\u7fd2\u524d\u50c5\u9700\u5c11\u91cf\u4eba\u985e\u77e5\u8b58\u7684\u4ecb\u5165\u3002\u795e\u7d93\u7db2\u8def\u662f\u7531\u591a\u500b\u795e\u7d93\u5143\u7d44\u6210\u7684\u6578\u5b78\u6a21\u578b\uff0c\u9019\u4e9b\u795e\u7d93\u5143\u6a21\u64ec\u4e86\u4eba\u8166\u795e\u7d93\u5143\u4e4b\u9593\u7684\u9023\u63a5\uff0c\u795e\u7d93\u7db2\u7d61\u901a\u5e38\u5305\u62ec\u8f38\u5165\u5c64\u3001\u96b1\u85cf\u5c64\u548c\u8f38\u51fa\u5c64\uff0c\u5176\u4e2d\u96b1\u85cf\u5c64\u7684\u6578\u91cf\u548c\u7d50\u69cb\u53ef\u4ee5\u6839\u64da\u4efb\u52d9\u9700\u6c42\u4f86\u8abf\u6574\uff0c\u4e4b\u6240\u4ee5\u7a31\u70ba\u6df1\u5ea6\u5b78\u7fd2\u5c31\u662f\u56e0\u70ba\u5b83\u7684\u795e\u7d93\u7db2\u7d61\u901a\u5e38\u5305\u542b\u6578\u5341\u5230\u6578\u767e\u500b\u96b1\u85cf\u5c64(\u65e9\u671f\u53d7\u9650\u65bc\u786c\u9ad4\u8a2d\u5099\uff0c\u985e\u795e\u7d93\u7db2\u8def\u7684\u96b1\u85cf\u5c64\u5f80\u5f80\u4e0d\u9054\u5341\u5c64)\uff0c\u7d2f\u7a4d\u7684\u795e\u7d93\u5143\u6578\u91cf\u53ef\u9ad8\u9054\u6578\u767e\u842c\u500b\uff0c\u9019\u7a2e\u6df1\u5ea6\u7d50\u69cb\u5141\u8a31\u795e\u7d93\u7db2\u7d61\u5b78\u7fd2\u5f9e\u6578\u64da\u4e2d\u63d0\u53d6\u4e0d\u540c\u5c64\u6b21\u7684\u7279\u5fb5\uff0c\u6709\u52a9\u65bc\u66f4\u597d\u5730\u7406\u89e3\u548c\u8655\u7406\u5927\u898f\u6a21\u3001\u9ad8\u7dad\u5ea6\u7684\u6578\u64da\u4ee5\u53ca\u8907\u96dc\u7684\u4efb\u52d9\uff0c\u7576\u7136\u5b83\u6240\u9700\u8981\u7684\u8a13\u7df4\u6578\u64da\u5c31\u76f8\u7576\u5927\u91cf\u3002\u4eba\u5de5\u667a\u6167\u6a21\u578b\u7684\u8a13\u7df4(training)\u904e\u7a0b\u5c31\u5982\u540c\u7269\u7406\u6a21\u578b\u7684\u7387\u5b9a(calibration)\u904e\u7a0b\uff0c\u900f\u904e\u7814\u7a76\u4eba\u54e1\u5c07\u6df1\u5ea6\u5b78\u7fd2\u6a21\u578b\u7684\u76ee\u6a19\u7269\u6a19\u7c64(labeling)\u63d0\u4f9b\u7d66\u6a21\u578b\uff0c\u901a\u5e38\u6703\u900f\u904e\u68af\u5ea6\u4e0b\u964d\u6cd5\u9019\u985e\u7684\u6700\u4f73\u5316\u65b9\u6cd5\uff0c\u8b93\u9810\u6e2c\u8207\u5be6\u969b\u7d50\u679c\u9593\u7684\u8aa4\u5dee\u6700\u5c0f\uff0c\u9010\u6f38\u8abf\u6574\u6a21\u578b\u7684\u53c3\u6578\u548c\u6b0a\u91cd\uff0c\u6700\u5f8c\u7372\u5f97\u4e00\u500b\u6700\u4f73\u5316\u7684\u6df1\u5ea6\u5b78\u7fd2\u6a21\u578b\uff0c\u56e0\u6b64\uff0c\u901a\u5e38\u6df1\u5ea6\u5b78\u7fd2\u6a21\u578b\u5728\u8a13\u7df4\u968e\u6bb5\u662f\u8017\u6642\u3001\u8017\u8a08\u7b97\u8cc7\u6e90\u7684\uff0c\u4f46\u5728\u57f7\u884c\u6642\u5247\u662f\u5feb\u901f\u7684\uff0c\u9019\u4e5f\u662f\u70ba\u4f55\u8fd1\u671fNvidia\u7684AI\u6676\u7247\u975e\u5e38\u53d7\u6b61\u8fce\uff0c\u56e0\u70ba\u5b83\u7684\u8a2d\u8a08\u8207\u74b0\u5883\u53ef\u4ee5\u52a0\u901f\u6a21\u578b\u7684\u8a13\u7df4\u3002<\/br><\/br>\nDeep learning, on the other hand, is based on neural network theory and differs from traditional machine learning methods by significantly reducing the need for human-knowledge-based feature extraction. It constructs complex neural networks using a large number of neurons and learns the hidden features within massive datasets through multi-layer neural network structures, requiring minimal human intervention before data learning. Neural networks are mathematical models composed of multiple neurons, simulating connections between human brain neurons. Typically, a neural network consists of an input layer, hidden layers, and an output layer, with the number and structure of hidden layers adjustable according to the task requirements. The term \"deep learning\" refers to the fact that such networks often contain dozens to hundreds of hidden layers (earlier constrained by hardware, neural networks rarely had more than ten layers), with accumulated neuron counts reaching millions. This deep structure allows neural networks to learn and extract hierarchical features from data, facilitating better understanding and handling of large-scale, high-dimensional data and complex tasks. Naturally, such networks require vast amounts of training data.<\/br><\/br>\nThe training process of AI models is akin to the calibration process in physical modeling, where researchers use optimization methods like gradient descent to minimize errors between predicted and actual results by adjusting model parameters and weights. As a result, the training phase of deep learning models is often time-consuming and computationally intensive, but they perform quickly during execution. This is one reason why Nvidia\u2019s AI chips are so popular recently, as their design and environment can accelerate model training.<\/br><\/br>\n\n\n\n\u4eba\u5de5\u667a\u6167\u900f\u904e\u89c0\u5bdf\u548c\u5206\u6790\u6578\u64da\u5f8c\uff0c\u53ef\u4ee5\u81ea\u52d5\u8b58\u5225\u3001\u627e\u51fa\u898f\u5f8b\u548c\u8da8\u52e2\uff0c\u4e26\u4f7f\u7528\u9019\u4e9b\u77e5\u8b58\u4f86\u505a\u51fa\u9810\u6e2c\u3001\u6c7a\u7b56\u548c\u89e3\u6c7a\u554f\u984c\uff0c\u6b32\u5b78\u7fd2\u7684\u554f\u984c\u672c\u8eab\u9700\u5b58\u5728\u67d0\u4e9b\u6f5b\u5728\u898f\u5247\u53ef\u4ee5\u53bb\u5b78\u7fd2\uff0c\u4e26\u4e14\u6709\u660e\u78ba\u7684\u76ee\u6a19\uff0c\u5b58\u5728\u3127\u5b9a\u898f\u5247\u537b\u76ee\u524d\u4e26\u7121\u7406\u8ad6\u516c\u5f0f\u53ef\u63cf\u8ff0\uff0c\u56e0\u6b64\u624d\u9808\u5229\u7528\u6a5f\u5668\u5b78\u7fd2\u6216\u6df1\u5ea6\u5b78\u7fd2\u4f86\u89e3\u6c7a\u554f\u984c\u3002\u800c\u6a5f\u5668\u5b78\u7fd2\u4ee5\u53ca\u6df1\u5ea6\u5b78\u7fd2\u4e2d\u63d0\u5230\u7684\u5b78\u7fd2\uff0c\u6307\u7684\u662f\u89c0\u5bdf\u4e00\u7d44\u8cc7\u6599\u4e26\u767c\u6398\u8cc7\u6599\u4e2d\u6f5b\u5728\u77e5\u8b58\u7684\u904e\u7a0b\uff0c\u5b78\u7fd2\u7684\u985e\u578b\u53c8\u53ef\u5206\u6210\u4e09\u985e\uff0c\u76e3\u7763\u5f0f\u5b78\u7fd2(supervised learning)\u7d66\u5b9a\u4e00\u7d44\u8cc7\u6599\u5305\u542b\u8981\u5224\u65b7\u6216\u9810\u6e2c\u7684\u7b54\u6848\u7d66\u6a5f\u5668\u9032\u884c\u5b78\u7fd2\u3001\u975e\u76e3\u7763\u5f0f\u5b78\u7fd2(unsupervised learning)\u7d66\u5b9a\u4e00\u7d44\u8cc7\u6599\u4f46\u4e0d\u5305\u542b\u8981\u5224\u65b7\u6216\u9810\u6e2c\u7684\u7b54\u6848\u4f86\u8b93\u6a5f\u5668\u5b78\u7fd2\u5c0d\u8cc7\u6599\u9032\u884c\u5206\u7fa4\u6216\u662f\u7d44\u7e54\u4ee5\u53ca\u5f37\u5316\u5f0f\u5b78\u7fd2(reinforcement learning)\u5247\u662f\u6a5f\u5668\u6703\u4e0d\u65b7\u5730\u5617\u8a66\u932f\u8aa4(trial-and-error)\u4e26\u91dd\u5c0d\u6f5b\u5728\u7684\u6b63\u78ba\u7b54\u6848\u5f97\u5230\u56de\u994b\u3002\u800c\u8fd1\u5e74\u56e0\u96fb\u8166\u8a08\u7b97\u80fd\u529b\u589e\u5f37\uff0c\u5c0d\u65bc\u5229\u7528\u8cc7\u6599\u63a2\u52d8\u3001\u6a5f\u5668\u5b78\u7fd2\u4ee5\u53ca\u6df1\u5ea6\u5b78\u7fd2\u7684\u7814\u7a76\u5c64\u51fa\u4e0d\u7aae\uff0c\u8b93\u6df1\u5ea6\u5b78\u7fd2\u884d\u751f\u51fa\u5404\u5f0f\u5404\u6a23\u7684\u529f\u80fd\uff0c\u6df1\u5ea6\u5b78\u7fd2\u5df2\u7d93\u904b\u7528\u5728\u975e\u5e38\u591a\u500b\u9818\u57df\uff0c\u5305\u62ec\u5716\u50cf\u8b58\u5225\u3001\u8a9e\u97f3\u8b58\u5225\u3001\u81ea\u7136\u8a9e\u8a00\u8655\u7406\u3001\u81ea\u52d5\u99d5\u99db\u3001\u91ab\u5b78\u5f71\u50cf\u5206\u6790\u3001\u91d1\u878d\u9818\u57df\u7b49\u3002\u4f8b\u5982\u81ea\u7136\u8a9e\u8a00\u8655\u7406(natural language processing, NLP)\u5247\u5c08\u6ce8\u5728\u4f7f\u96fb\u8166\u80fd\u7406\u89e3\u3001\u751f\u6210\u548c\u8655\u7406\u4eba\u985e\u81ea\u7136\u8a9e\u8a00\u7684\u9818\u57df\uff0c\u6db5\u84cb\u8a9e\u8a00\u6a21\u578b\u3001\u6587\u672c\u5206\u6790\u3001\u6a5f\u5668\u7ffb\u8b6f\u7b49\u65b9\u9762\uff1b\u5f37\u5316\u5b78\u7fd2(reinforcement learning)\u900f\u904e\u8a66\u8aa4\u6cd5\u4f86\u8a13\u7df4\u4ee3\u7406(\u5982\u6a5f\u5668\u4eba\u6216\u904a\u6232\u73a9\u5bb6)\u4ee5\u9054\u5230\u6700\u4f73\u5316\u67d0\u5c0d\u8c61\u70ba\u76ee\u6a19\uff1b\u96fb\u8166\u8996\u89ba(computer vision)\u81f4\u529b\u65bc\u4f7f\u8a08\u7b97\u6a5f\u80fd\u5920\u7406\u89e3\u548c\u89e3\u91cb\u8996\u89ba\u8cc7\u8a0a\uff0c\u542b\u62ec\u7167\u7247\u548c\u5f71\u7247\u7684\u5206\u6790\u3001\u7269\u4ef6\u5075\u6e2c\u548c\u5834\u666f\u7406\u89e3\u7b49\u3002<\/br><\/br>\n\nAfter observing and analyzing data, AI can automatically identify patterns and trends and use this knowledge to make predictions, decisions, and solve problems. The problem to be learned must have inherent patterns to be discovered and a clear objective that is difficult to describe using theoretical formulas, making machine learning or deep learning necessary for problem-solving. The learning process in machine learning and deep learning refers to discovering latent knowledge by observing a dataset. Learning can be classified into three types: supervised learning, which involves providing a dataset containing answers to be predicted or identified for the machine to learn; unsupervised learning, which involves providing a dataset without predefined answers to allow the machine to perform clustering or organization; and reinforcement learning, where the machine continuously tries and errors and receives feedback based on potential correct answers.<\/br><\/br>\n\n\nWith the increase in computing power in recent years, research using data mining, machine learning, and deep learning has become increasingly prevalent, leading to the development of various functions derived from deep learning. Deep learning has been applied in numerous fields, including image recognition, speech recognition, natural language processing, autonomous driving, medical image analysis, and finance. For example, natural language processing (NLP) focuses on enabling computers to understand, generate, and process human natural language, covering areas such as language modeling, text analysis, and machine translation. Reinforcement learning trains agents (such as robots or game players) to optimize specific objectives through trial and error. Computer vision strives to enable computers to understand and interpret visual information, encompassing image and video analysis, object detection, and scene understanding.<\/br><\/br>\n\n\u591a\u985e\u5225\u7269\u4ef6\u5206\u985e\u8207\u5075\u6e2c(multiple object classification and detection)\u662f\u96fb\u8166\u8996\u89ba\u9818\u57df\u5e38\u898b\u7684\u4efb\u52d9\u4e4b\u4e00\uff0c\u53ef\u4ee5\u505a\u5230\u5982\u5f71\u50cf\u5206\u985e(image classification)\u4f86\u5224\u65b7\u5f71\u50cf\u4e2d\u6709\u5e7e\u7a2e\u7269\u4ef6\uff0c\u4e5f\u53ef\u4ee5\u505a\u5230\u5982\u7269\u4ef6\u5b9a\u4f4d(object localization)\u53ef\u4ee5\u5224\u65b7\u5f71\u50cf\u4e2d\u7684\u7269\u4ef6\u9084\u80fd\u5b9a\u4f4d\u7269\u4ef6\u5728\u756b\u9762\u4e2d\u7684\u4f4d\u7f6e\u3002\u800c\u4eba\u5de5\u667a\u6167\u6280\u8853\u5728\u96fb\u8166\u8996\u89ba\u9818\u57df\u7684\u6210\u6548\u80fd\u6b78\u529f\u65bc\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\u4e2d\u7684\u5377\u7a4d\u795e\u7d93\u7db2\u8def(convolutional neural network, CNN)\u7684\u767c\u5c55\uff0cCNN\u80fd\u5920\u6709\u6548\u5730\u5728\u5f71\u50cf\u4e2d\u63d0\u53d6\u4e8c\u7dad\u7279\u5fb5\uff0c\u5206\u6790\u7279\u5fb5\u4f86\u63a8\u8ad6\u5f71\u50cf\u4e2d\u7684\u5167\u5bb9\u3002\u7136\u800c\uff0c\u4f7f\u7528CNN\u50c5\u80fd\u9032\u884c\u5f71\u50cf\u5206\u985e\u4efb\u52d9\uff0c\u7121\u6cd5\u5c07\u6240\u6709\u7684\u7269\u4ef6\u8fa8\u8a8d\u51fa\u4f86\u4e26\u4e14\u6a19\u793a\u7269\u4ef6\u5728\u756b\u9762\u4e2d\u7684\u4f4d\u7f6e(\u4f8b\u5982\u7528\u77e9\u5f62\u6846\u6a19\u793a)\u3002\u56e0\u6b64\uff0c\u6839\u64daCNN\u7684\u7406\u8ad6\u57fa\u790e\uff0c\u6709\u7814\u7a76\u63d0\u51faR-CNN(Region-CNN)\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\uff0c\u900f\u904e\u5728\u5f71\u50cf\u4e2d\u5283\u5206\u51fa\u591a\u500b\u5019\u9078\u5340\u57df(region proposal)\uff0c\u5c0d\u6bcf\u500b\u5019\u9078\u5340\u57df\u9032\u884c\u5377\u7a4d\u904e\u7a0b\uff0c\u5224\u65b7\u756b\u9762\u4e2d\u7684\u7269\u4ef6\u4f4d\u7f6e\u8207\u985e\u5225\u3002R-CNN\u7684\u63a8\u51fa\u5960\u5b9a\u4e86\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\u5728\u7269\u4ef6\u5075\u6e2c\u4efb\u52d9\u4e0a\u7684\u57fa\u790e\uff0c\u540c\u6a23R-CNN\u7684\u7814\u7a76\u5718\u968a\u5f88\u5feb\u5730\u63a8\u51fa\u4e86\u66f4\u65b0\u7248\u672c\u7684Fast R-CNN\uff0c\u52a0\u5165RoI\u6c60\u5316\u5c64(region of interest pooling)\uff0c\u5c07\u6bcf\u500b\u5019\u9078\u5340\u57df\u5148\u9032\u884c\u6c60\u5316\u518d\u5377\u7a4d\uff0c\u5f97\u5230\u56fa\u5b9a\u5c3a\u5bf8\u7684\u7279\u5fb5\u5716\uff0c\u80fd\u6e1b\u5c11\u904b\u7b97\u904e\u7a0b\u4e26\u4e14\u4fdd\u7559\u5f71\u50cf\u7279\u5fb5\uff0c\u63d0\u5347\u904b\u7b97\u901f\u5ea6\u3002\u6709\u7814\u7a76\u8a8d\u70baFast R-CNN\u7684\u904b\u7b97\u901f\u5ea6\u9084\u80fd\u518d\u63d0\u9ad8\uff0c\u63a8\u51faFaster R-CNN\uff0c\u52a0\u5165\u5340\u57df\u63d0\u53d6\u7db2\u8def(region proposal network, RPN)\u4f86\u53d6\u4ee3\u539f\u5148Fast R-CNN\u5283\u5206\u51fa\u5019\u9078\u5340\u57df\u7684\u65b9\u5f0f\uff0c\u5c07\u9019\u4e9b\u7db2\u8def\u7522\u751f\u7684\u5019\u9078\u5340\u57df\u7a31\u70ba\u9328(anchors)\uff0c\u5c07\u5019\u9078\u5340\u57df\u7684\u9078\u64c7\u3001\u5f71\u50cf\u7279\u5fb5\u7684\u63d0\u53d6\u4ee5\u53ca\u7269\u4ef6\u5206\u985e\u8207\u5b9a\u4f4d\u7b49\u4efb\u52d9\u8a2d\u8a08\u6210\u540c\u500b\u7db2\u8def\u67b6\u69cb\u4f86\u8a13\u7df4\uff0c\u6709\u6548\u6e1b\u5c11\u904b\u7b97\u6642\u9593\u4e26\u63d0\u5347\u901f\u5ea6\u3002<\/br><\/br>\n\n\n\n\nMulti-object classification and detection is a common task in the field of computer vision, capable of tasks such as image classification to identify the number of objects in an image or object localization to determine the position of objects within an image. The success of AI technology in computer vision can be attributed to the development of convolutional neural networks (CNNs) within deep learning methods. CNNs can efficiently extract two-dimensional features from images and use these features to infer image content. However, CNNs can only perform image classification tasks and cannot identify and localize all objects in the image (such as marking them with bounding boxes). Therefore, based on the theoretical foundation of CNNs, researchers proposed the R-CNN (Region-CNN) deep learning method, which divides an image into multiple candidate regions (region proposals) and performs convolution on each region to determine the location and class of objects in the image.<\/br><\/br>\n\nThe introduction of R-CNN laid the groundwork for deep learning methods in object detection tasks. The same research team quickly introduced an updated version, Fast R-CNN, which added the Region of Interest (RoI) pooling layer to pool each candidate region before convolution, obtaining fixed-size feature maps, reducing computation time while preserving image features and improving processing speed. Researchers believed that the computation speed of Fast R-CNN could be further improved, leading to the introduction of Faster R-CNN, which incorporated a Region Proposal Network (RPN) to replace the original method of generating candidate regions in Fast R-CNN. The candidate regions produced by this network were referred to as anchors, and the task of selecting candidate regions, extracting image features, and object classification and localization were designed into a single network architecture for training, effectively reducing computation time and increasing speed.<\/br><\/br>\n\n\u96d6\u7136R-CNN\u7684\u51fa\u73fe\u5960\u5b9a\u4e86\u7269\u4ef6\u5075\u6e2c\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\u7684\u57fa\u790e\uff0c\u4f46\u6709\u7814\u7a76\u8a8d\u70ba\u5982R-CNN\u9019\u7a2e\u5169\u968e\u6bb5\u5b78\u7fd2\u65b9\u6cd5(two-stage learning)\u5728\u904b\u7b97\u6548\u7387\u63d0\u5347\u4e0a\u8f03\u96e3\u4ee5\u7a81\u7834\uff0c\u56e0\u6b64\uff0c\u63a8\u51faYOLO(you only look once)\u6f14\u7b97\u6cd5\uff0c\u5f37\u8abf\u5176\u904b\u7b97\u6548\u7387\uff0c\u5c07\u7269\u4ef6\u4f4d\u7f6e\u5075\u6e2c\u8207\u8fa8\u8b58\u8996\u70ba\u55ae\u4e00\u7684\u56de\u6b78\u554f\u984c\uff0c\u5f9e\u5f71\u50cf\u8f38\u5165\u5230\u8f38\u51fa\u7d50\u679c\u50c5\u4f9d\u9760\u4e00\u500b\u795e\u7d93\u7db2\u8def\u4f86\u904b\u7b97\uff0c\u5728\u8a13\u7df4\u6642\u5c07\u6574\u5f35\u5f71\u50cf\u4f5c\u70ba\u5b78\u7fd2\u7684\u76ee\u6a19\uff0cYOLO\u6f14\u7b97\u6cd5\u5728\u8a13\u7df4\u8207\u8fa8\u8b58\u6642\uff0c\u5c0d\u540c\u5f35\u5f71\u50cf\u50c5\u300c\u770b\u300d\u4e00\u6b21\uff0c\u5373\u5176\u547d\u540d\u7684\u7531\u4f86\u3002\u53e6\u5916\uff0c\u540c\u6a23\u70ba\u4e00\u968e\u6bb5\u5b78\u7fd2\u65b9\u6cd5\u7684SSD(single shot multibox detector)\u6f14\u7b97\u6cd5\uff0c\u4e5f\u662f\u5c07\u7269\u4ef6\u4f4d\u7f6e\u5075\u6e2c\u8207\u7269\u4ef6\u985e\u5225\u8fa8\u8b58\u5408\u4f75\u6210\u4e00\u500b\u795e\u7d93\u7db2\u8def\u5c31\u53ef\u4ee5\u5b8c\u6210\u6240\u6709\u7684\u591a\u7269\u4ef6\u5206\u985e\u8207\u5075\u6e2c\u4efb\u52d9\uff0cSSD\u63a1\u7528\u6574\u500bCNN\u7db2\u8def\u9032\u884c\u4f4d\u7f6e\u5075\u6e2c\u8207\u8fa8\u8b58\uff0c\u4e26\u4e14\u904b\u7b97\u6642\u63d0\u53d6\u4e0d\u540c\u5c3a\u5ea6\u7684\u7279\u5fb5\u5716\uff0c\u5f37\u5316\u5728\u5927\u6216\u5c0f\u7269\u4ef6\u4e0a\u7684\u8fa8\u8b58\u6548\u679c\u3002\u591a\u7269\u4ef6\u5206\u985e\u8207\u5075\u6e2c\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\u767c\u5c55\u521d\u671f\uff0c\u4ee5\u4e0a\u8ff0\u4e09\u7a2e\u6f14\u7b97\u6cd5\u851a\u70ba\u4e3b\u6d41\uff0c\u5728\u6e96\u78ba\u6027\u8207\u904b\u7b97\u6548\u7387\u4e0a\u5404\u6709\u512a\u52a3\uff0c\u5728\u76f8\u540c\u7684\u61c9\u7528\u4efb\u52d9\u4e0b\uff0cFaster R-CNN\u901a\u5e38\u53ef\u4ee5\u7372\u5f97\u8f03\u9ad8\u6e96\u78ba\u6027\u7684\u7d50\u679c\uff0c\u4f46\u904b\u7b97\u6548\u7387\u9060\u4e0d\u53ca\u5176\u4ed6\u5169\u7a2e\u6f14\u7b97\u6cd5\uff0cYOLO\u8207SSD\u7686\u80fd\u4ee5\u5373\u6642\u904b\u7b97\u901f\u5ea6\u9ad8\u6548\u904b\u7b97\uff0c\u5176\u4e2dSSD\u53c8\u6bd4YOLO\u5feb\u4e00\u4e9b\uff0c\u4f46YOLO\u6f14\u7b97\u6cd5\u4ee5\u8f03\u4f73\u7684\u8fa8\u8b58\u6e96\u78ba\u6027\u4f54\u512a\u3002<\/br><\/br>\n\nAlthough the introduction of R-CNN established the foundation for deep learning methods in object detection, researchers found that two-stage learning methods like R-CNN were challenging to improve in terms of computational efficiency. Consequently, the YOLO (You Only Look Once) algorithm was introduced, emphasizing computational efficiency by viewing object localization and recognition as a single regression problem. From image input to output, a single neural network is used for computation. During training, the entire image is used as the learning target, and the YOLO algorithm processes each image only once, hence its name. Similarly, the SSD (Single Shot MultiBox Detector) algorithm, also a single-stage learning method, combines object localization and classification into a single neural network capable of completing all multi-object classification and detection tasks. SSD employs an entire CNN network for localization and recognition, extracting feature maps at different scales to enhance recognition performance for large and small objects.<\/br><\/br>\n\n\nYOLO\u6f14\u7b97\u6cd5\u5c07\u7269\u4ef6\u5075\u6e2c\u8207\u5206\u985e\u7684\u4efb\u52d9\u6574\u5408\u5230\u55ae\u4e00\u7684\u7db2\u8def\u4e2d\uff0c\u7a31\u70ba\u7d71\u4e00\u5075\u6e2c(unified detection)\u3002\u5176\u5de5\u4f5c\u539f\u7406\u662f\u5148\u5c07\u5f71\u50cf\u5283\u5206\u6210\u591a\u500b\u7db2\u683c\uff0c\u7136\u5f8c\u91dd\u5c0d\u6bcf\u500b\u7db2\u683c\u4e2d\u7684\u8cc7\u8a0a\u9032\u884c\u904b\u7b97\uff0c\u5f9e\u6574\u5f35\u5f71\u50cf\u4e2d\u63d0\u53d6\u7279\u5fb5\u4e26\u8fa8\u8b58\u7269\u4ef6\u3002\u9019\u6a23\u7684\u7db2\u8def\u7d50\u69cb\u53ef\u5f9e\u6574\u5f35\u5f71\u50cf\u4e2d\u63d0\u53d6\u7279\u5fb5\uff0c\u4e26\u4e14\u7528\u4f86\u8fa8\u8b58\u76ee\u6a19\u7269\uff0c\u540c\u6642\u5728\u6574\u5f35\u5f71\u50cf\u4e0a\u91dd\u5c0d\u4e0d\u540c\u7684\u76ee\u6a19\u7269\u4ee5\u908a\u754c\u6846\u6a19\u793a\u5176\u4f4d\u7f6e\u8207\u985e\u5225\u3002\u9019\u4f7f\u5f97\u7db2\u8def\u53ef\u4ee5\u5206\u6790\u6574\u5f35\u5f71\u50cf\u7684\u7279\u5fb5\uff0c\u7dad\u6301\u4e00\u5b9a\u7684\u8fa8\u8b58\u80fd\u529b\u4e26\u63d0\u9ad8\u8fa8\u8b58\u901f\u5ea6\u3002\u5728YOLO\u6f14\u7b97\u6cd5\u4e2d\uff0c\u5f71\u50cf\u88ab\u5283\u5206\u6210\ud835\udc46\u00d7\ud835\udc46\u7684\u7db2\u683c\u3002\u5982\u679c\u76ee\u6a19\u7269\u7684\u4e2d\u5fc3\u4f4d\u65bc\u67d0\u500b\u7db2\u683c\u5167\uff0c\u5247\u6b64\u7db2\u683c\u8ca0\u8cac\u6b64\u76ee\u6a19\u7269\u7684\u5075\u6e2c\u3002\u6bcf\u500b\u7db2\u683c\u53ef\u8f38\u51fa\ud835\udc35\u500b\u908a\u754c\u6846\u53ca\u5176\u5404\u81ea\u7684\u4fe1\u5fc3\u5206\u6578(confidence score)\u3002<\/br><\/br>\n\n\nIn the early stages of multi-object classification and detection deep learning methods, the aforementioned three algorithms were dominant, each with its strengths and weaknesses in terms of accuracy and computational efficiency. For the same application task, Faster R-CNN typically achieved higher accuracy but was far less efficient than the other two algorithms. Both YOLO and SSD could operate with real-time processing speed, with SSD being slightly faster but YOLO achieving better recognition accuracy.\nThe YOLO algorithm integrates object detection and classification into a single network, known as unified detection. Its working principle is to first divide the image into multiple grids and then process the information in each grid, extracting features and recognizing objects from the entire image. This network structure extracts features from the whole image, recognizes target objects, and simultaneously marks their locations and categories in the image using bounding boxes. This allows the network to analyze features from the entire image, maintaining recognition capability and improving recognition speed. In the YOLO algorithm, the image is divided into an S\u00d7S grid. If the center of an object is within a particular grid, that grid is responsible for detecting that object. Each grid outputs B bounding boxes and their respective confidence scores.<\/br><\/br>\n\n\nYOLO\u5229\u7528\u55ae\u500b\u5377\u7a4d\u795e\u7d93\u7db2\u8def\u67b6\u69cb\u5efa\u7f6e\u8fa8\u8b58\u6a21\u578b\u4f86\u9032\u884c\u591a\u7269\u4ef6\u5075\u6e2c\u8207\u8fa8\u8b58\uff0cYOLO\u7684\u795e\u7d93\u7db2\u8def\u67b6\u69cb\u7531\u6700\u4e00\u958b\u59cb\u7684\u5377\u7a4d\u5c64\u8ca0\u8cac\u5f9e\u8f38\u5165\u5f71\u50cf\u4e2d\u63d0\u53d6\u7279\u5fb5\uff0c\u5b8c\u5168\u9023\u63a5\u5c64\u5247\u8ca0\u8cac\u8f38\u51fa\u76ee\u6a19\u7269\u7684\u985e\u5225\u8207\u908a\u754c\u6846\u7684\u6a5f\u7387\u8207\u5ea7\u6a19\u3002YOLO\u53d7\u5230GoogLeNet\u6a21\u578b\u7684\u555f\u767c\uff0c\u7db2\u8def\u7d50\u69cb\u4e2d\u5305\u542b24\u5c64\u5377\u7a4d\u5c64\u4ee5\u53ca2\u500b\u5168\u9023\u63a5\u5c64\uff0c\u4e0d\u540c\u4e4b\u8655\u5728\u65bcYOLO\u7c21\u55ae\u5730\u4f7f\u75281*1\u7684\u7e2e\u6e1b\u5c64(reduction layers)\u82073*3\u7684\u5377\u7a4d\u5c64\u53d6\u4ee3GoogLeNet\u6240\u4f7f\u7528\u7684Inception\u6a21\u7d44\u3002\u5168\u9023\u63a5\u5c64\u6703\u8f38\u51fa\u6bcf\u4e00\u500b\u5305\u542b\u76ee\u6a19\u7269\u7684\u908a\u754c\u6846\u985e\u5225\u6a5f\u7387\u8207\u5176\u5ea7\u6a19\u4f4d\u7f6e\uff0cYOLO\u5c07\u8f38\u51fa\u908a\u754c\u6846\u7684\u9577\u8207\u5bec\u91dd\u5c0d\u6574\u5f35\u5f71\u50cf\u7684\u9577\u8207\u5bec\u6b63\u898f\u5316\u52300\u81f31\u7684\u7bc4\u570d\uff0c\u908a\u754c\u6846\u7684\u4e2d\u5fc3\u9ede\u5ea7\u6a19\u4e5f\u540c\u6642\u6b63\u898f\u5316\u52300\u81f31\u7684\u7bc4\u570d\u3002\u800c\u5728\u5404\u500b\u795e\u7d93\u7db2\u8def\u5c64\u7684\u555f\u52d5\u51fd\u6578\u4e2d\uff0c\u9664\u4e86\u5728\u6700\u5f8c\u4e00\u5c64\u4f7f\u7528\u7dda\u6027\u555f\u52d5\u51fd\u6578(linear activation function)\u5916\uff0c\u5176\u9918\u7686\u4f7f\u7528\u6d29\u6f0fReLU\u555f\u52d5\u51fd\u6578(leaky ReLU activation function)\u3002 \n\u7269\u4ef6\u5075\u6e2c\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\u5728\u8a13\u7df4\u6642\uff0c\u662f\u900f\u904e\u4e0d\u65b7\u8fed\u4ee3\u8a13\u7df4\u8abf\u6574\u5167\u90e8\u53c3\u6578\uff0c\u5617\u8a66\u964d\u4f4e\u6a21\u578b\u7684\u8aa4\u5dee\u76f4\u5230\u6700\u4f73\uff0c\u56e0\u6b64\u8a08\u7b97\u8aa4\u5dee\u7684\u65b9\u5f0f\u6703\u5f71\u97ff\u6a21\u578b\u7684\u8a13\u7df4\u6548\u679c\uff0c\u5373\u8a08\u7b97\u8aa4\u5dee\u7684\u640d\u5931\u51fd\u6578(loss function)\u5341\u5206\u91cd\u8981\u3002YOLO\u6f14\u7b97\u6cd5\u81ea\u554f\u4e16\u4ee5\u4f86\uff0c\u5df2\u7d93\u63a8\u51fa\u4e86\u591a\u500b\u7248\u672c\uff0c\u8fd1\u671f\u4ee5YOLOv7\u70ba\u5ee3\u6cdb\u4f7f\u7528\uff0c\u7121\u8ad6\u662f\u5728\u8a13\u7df4\u904e\u7a0b\u8207\u5be6\u969b\u4f7f\u7528\u4e0a\u90fd\u8868\u73fe\u51fa\u9ad8\u5ea6\u7684\u7a69\u5b9a\u6027\u8207\u6e96\u78ba\u6027\u3002\u6b64\u5916\uff0cYOLOv7\u7684\u958b\u767c\u5718\u968a\u9084\u7121\u511f\u63d0\u4f9b\u4e86\u958b\u6e90\u4e14\u6613\u65bc\u4fee\u6539\u7684\u7a0b\u5f0f\u78bc\uff0c\u53ef\u4ee5\u6839\u64da\u8a08\u756b\u7684\u4e0d\u540c\u9700\u6c42\u9032\u884c\u9748\u6d3b\u4fee\u6539\u3002\u53f0\u5357\u5e02\u7684\u667a\u6167\u6d77\u7058\u5075\u6e2c\u7cfb\u7d71\u4ea6\u662f\u4f7f\u7528YOLOv7\u6f14\u7b97\u6cd5\u4f86\u5efa\u7f6e\uff0c\u8b93AI\u5b78\u7fd2\u5982\u4f55\u8fa8\u8b58\u6d77\u57df\u904a\u61a9\u6d3b\u52d5\u8207\u6d77\u7058\u4e0a\u7684\u6c11\u773e\uff0c\u4e5f\u80fd\u5b78\u7fd2\u5982\u4f55\u8fa8\u8b58\u53ef\u80fd\u767c\u751f\u6eba\u6c34\u7684\u6c11\u773e\uff0c\u751a\u81f3\u5b78\u7fd2\u96e2\u5cb8\u6d41\u767c\u751f\u6642\u7684\u7279\u5fb5\u3002<\/br><\/br>\n\nYOLO uses a single convolutional neural network architecture to construct a recognition model for multi-object detection and recognition. The YOLO network structure consists of 24 convolutional layers and two fully connected layers. Unlike GoogLeNet, which uses the Inception module, YOLO simply employs 1\u00d71 reduction layers and 3\u00d73 convolutional layers. The fully connected layers output the class probabilities and coordinates for each object bounding box, with the bounding box's width and height normalized to the image dimensions between 0 and 1. The center coordinates of the bounding box are also normalized to the range of 0 to 1. In the activation functions for each neural network layer, all except the last layer use the leaky ReLU activation function, while the last layer uses a linear activation function.<\/br><\/br>\n\n\nDuring the training of object detection deep learning methods, the internal parameters are iteratively adjusted to minimize the model's error until it is optimized. Thus, the method for calculating errors, the loss function, is crucial to the model's training effectiveness. Since its introduction, the YOLO algorithm has released multiple versions, with YOLOv7 being widely used recently. It demonstrates high stability and accuracy in both training and practical use. Additionally, the YOLOv7 development team has provided open-source and easily modifiable code, allowing flexible adjustments according to the needs of different projects. The smart beach detection system in Tainan City also uses the YOLOv7 algorithm, enabling AI to learn how to identify recreational activities and people on the beach, as well as potentially recognizing drowning individuals and detecting features of rip currents.<\/br><\/br>\n\n<\/h5>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t
\n\n\n\n<\/h5>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t
YOLO\u6f14\u7b97\u6cd5\u4f7f\u7528\u7d71\u4e00\u5075\u6e2c(unified detection)\u4f86\u8fa8\u8b58\u76ee\u6a19\u7269\u7684\u4f4d\u7f6e\u8207\u985e\u5225<\/br>\nThe YOLO algorithm uses unified detection to identify the location and class of target \n<\/h5>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t
\u6d77\u57df\u904a\u61a9\u6d3b\u52d5\u8207\u4eba\u6d41\u76e3\u6e2c\u6a21\u578b\u5373\u6642\u8fa8\u8b58\u8207\u7d71\u8a08\u756b\u9762\u3002\u904b\u7528AI\u6280\u8853\u53ef\u4ee5\u5c07\u756b\u9762\u4e2d\u7684\u904a\u61a9\u6c11\u773e\u4ee5\u4e0d\u540c\u984f\u8272\u8207\u6d3b\u52d5\u540d\u7a31\u7684\u908a\u754c\u6846\u6a19\u793a\u5176\u4f4d\u7f6e\u8207\u985e\u5225\uff0c\u76e3\u6e2c\u7d50\u679c\u4e5f\u80fd\u7d93\u904e\u7d71\u8a08\u5f8c\u986f\u793a\u73fe\u5834\u6d3b\u52d5\u4eba\u6578\u6982\u6cc1 (\u6210\u679c\u6848\u4f8b\u6458\u81ea: \u667a\u6167\u5316\u6d77\u57df\u74b0\u5883\u76e3\u6e2c\u7cfb\u7d71\u5efa\u7f6e\u4e4b\u7814\u7a76-\u6c34\u57df\u904a\u61a9\u6d3b\u52d5\u5b89\u5168\u76e3\u6e2c\u6280\u8853\u7814\u767c\uff0c\u570b\u5bb6\u6d77\u6d0b\u7814\u7a76\u9662\uff0c2023)<\/br>\nReal-time recognition and statistical display of the coastal recreational activities and crowd monitoring model. Using AI technology, recreational participants in the scene can be marked with bounding boxes of different colors and activity names to indicate their location and category. The monitoring results can also be statistically analyzed to display an overview of the number of people engaged in various activities on site<\/h5>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t
\u6a21\u578b\u5075\u6e2c\u6f5b\u5728\u6eba\u6c34\u6c11\u773e\u3002\u7576\u53ef\u80fd\u6709\u6c11\u773e\u767c\u751f\u6eba\u6c34\u6642\uff0c\u6a21\u578b\u4ee5\u8b66\u793a\u7528\u7d05\u8272\u6a19\u793a\u6c11\u773e\u4e26\u767c\u51fa\u8b66\u544a\uff0c\u540c\u6642\u53c8\u4e0d\u6703\u5c07\u5176\u4ed6\u6b63\u5e38\u6d3b\u52d5\u8aa4\u5831\u6210\u6eba\u6c34 (\u6210\u679c\u6848\u4f8b\u6458\u81ea: \u667a\u6167\u5316\u6d77\u57df\u74b0\u5883\u76e3\u6e2c\u7cfb\u7d71\u5efa\u7f6e\u4e4b\u7814\u7a76-\u6c34\u57df\u904a\u61a9\u6d3b\u52d5\u5b89\u5168\u76e3\u6e2c\u6280\u8853\u7814\u767c\uff0c\u570b\u5bb6\u6d77\u6d0b\u7814\u7a76\u9662\uff0c2023)<\/br>\nThe model detects potential drowning individuals. When a person is likely to be drowning, the model highlights the individual with a red warning marker and issues an alert, while avoiding false positives for normal activities.\n\n<\/h5>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t
\u6f5b\u5728\u6eba\u6c34\u8207\u8de8\u8d8a\u8b66\u6212\u7dda\u8ffd\u8e64\u6a21\u578b\u5373\u6642\u8fa8\u8b58\u8207\u7d71\u8a08\u756b\u9762\u3002\u904b\u7528\u7269\u4ef6\u5075\u6e2c\u6df1\u5ea6\u5b78\u7fd2\u65b9\u6cd5\u8207\u591a\u76ee\u6a19\u8ffd\u8e64\u6280\u8853\uff0c\u80fd\u5920\u8ffd\u8e64\u53ef\u80fd\u6eba\u6c34\u6216\u8de8\u8d8a\u8b66\u6212\u7dda\u7684\u904a\u61a9\u6c11\u773e (\u6210\u679c\u6848\u4f8b\u6458\u81ea: \u667a\u6167\u5316\u6d77\u57df\u74b0\u5883\u76e3\u6e2c\u7cfb\u7d71\u5efa\u7f6e\u4e4b\u7814\u7a76-\u6c34\u57df\u904a\u61a9\u6d3b\u52d5\u5b89\u5168\u76e3\u6e2c\u6280\u8853\u7814\u767c\uff0c\u570b\u5bb6\u6d77\u6d0b\u7814\u7a76\u9662\uff0c2023)<\/br>\nReal-time recognition and statistical display of the potential drowning and boundary-crossing tracking model. Utilizing object detection deep learning methods and multi-target tracking technology, the model can track recreational participants who may be drowning or crossing warning lines\n<\/h5>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\n\t\t\t\t\t\t<\/span>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t
\u6b64\u70ba\u53f0\u5357\u5e02\u667a\u6167\u6d77\u7058\u6574\u7406\u898f\u5283\u6848\u4e4b\u793a\u7bc4\u7ad9\u7db2\u9801<\/br>\n\u7d93\u8cbb\u88dc\u52a9\uff1a\u6d77\u6d0b\u59d4\u54e1\u6703<\/br>\n\u6307\u5c0e\uff1a\u5e02\u653f\u5e9c\u89c0\u5149\u65c5\u904a\u5c40<\/br>\n\u57f7\u884c\u8207\u5efa\u7f6e\uff1a\u570b\u7acb\u6210\u529f\u5927\u5b78\u8fd1\u6d77\u6c34\u6587\u4e2d\u5fc3<\/br>\n\u7814\u7a76\u671f\u7a0b\uff1a113\u5e746\u6708\u81f311\u6708\n<\/h5>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\n\t\t\t\t\t\t<\/span>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"

\u4eba\u5de5\u667a\u6167(artificial inte […]<\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"site-sidebar-layout":"no-sidebar","site-content-layout":"","ast-site-content-layout":"full-width-container","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"disabled","ast-breadcrumbs-content":"","ast-featured-img":"disabled","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"class_list":["post-51","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/pages\/51","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/comments?post=51"}],"version-history":[{"count":17,"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/pages\/51\/revisions"}],"predecessor-version":[{"id":337,"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/pages\/51\/revisions\/337"}],"wp:attachment":[{"href":"https:\/\/www.comc.ncku.edu.tw\/smartbeach\/wp-json\/wp\/v2\/media?parent=51"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}