【案例】百度千帆大模型在工业智能制造领域的应用

微信扫码，打开到小程序

不畏浮云遮望眼 Baidu Industry Empowerment Collection 2023 Autonomous and Controllable AI Industry enabling platform China’s The first fully open source, fully functional industrial level deep-learning platform The world’s leading self- driving solutions Open, complete, and secure autonomous driving software and hardware integrated solutions, open platform, and ecosystem Kun-Lun Hong-Hu China’s first full-function AI chip with Highest Power/High Cost-Effective/Easy to Use Far-field voice interaction chip, automotive- grade standard, ultra-large memory, low power consumption Tian-Suan Data Lake analysis platform Tian-Gong Edge convergence IOT Platform Tian-He Cloud native development platform Tian-Lian Cloud area platform Tian-Xiang Intelligent multimedia platform Kai-Wu Industrial Internet Platform # encoding:utf-8 import cv2 face_cascade = cv2.CascadeClassifier('haarcascade_files/haarcascade_frontalface_default.xml') eye_cascade = cv2.CascadeClassifier('haarcascade_files/haarcascade_eye.xml') # 读取图像 img = cv2.imread('west.jpeg') gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 转为灰度图 # 检测脸部 faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5, minSize=(30, 30), flags=cv2.CASCADE_SCALE_IMAGE) print('Detected ', len(faces), " face") # 标记位置 for (x, y, w, h) in faces: img = cv2.rectangle(img, (x, y), (x + w, y + h), (255, 0, 0), 1) # cv2.circle(img, (int((x + x + w) / 2), int((y + y + h) / 2)), int(w / 2), (0, 255, 0), 1) roi_gray = gray[y: y + h, x: x + w] roi_color = img[y: y + h, x: x + w] eyes = eye_cascade.detectMultiScale(roi_gray) for (ex, ey, ew, eh) in eyes: cv2.rectangle(roi_color, (ex, ey), (ex + ew, ey + eh), (0, 255, 0), 1) label = 'Result: Detected ' + str(len(faces)) +" faces !" cv2.putText(img, label, (10, 20), cv2.FONT_HERSHEY_SCRIPT_COMPLEX, 0.8, (0, 0, 0), 1) # 显示图像 cv2.imshow('img', img) cv2.waitKey(0) cv2.destroyAllWindows() Haar Feature Extraction Adaboost Training OpenCV Environmental Installation Image Integration Image Reating 10000+ Data Acquisition Recursive Operation Matrix Operation Face Recognition Coordinate Box Selection 1000000+ Data Training import paddlehub as hub module = hub.Module(name="ultra_light_fast_generic_face_detector_1 mb_640") res = module.face_detection(paths = ["./test.jpg"], visualization=True, output_dir='face_detection_output') Vocational Colleges 【Scientific Research Cooperation】 Ecological enterprise 【Project Cooperation】 Requirement Scenario 【Technological Empowerment】 Intelligent Manufacturing Workshop TianGong/KaiWu/Smart Production Line/ Industrial Vision/Electrical Automation/ Industrial Control Unit Industrial Research Center Computing Power Cluster/BML/EasyDL EasyData/EasyEdge/UNIT Talent Innovation Training Center Operation & Service System AI Industry Development Center Industrial Intelligence Intelligent Industrialization Industry Development Center Eco-Enterprises Demand Side Professional Technicians AI Industry Crowdsourcing Platform Obtain Project Opportunities｜Supply-Demand Information Connecting｜ Project Management AI Talent Community and Precision Service Platform Industry-education integration training | professional talent connecting | technical exchange community AI Industry Standardised Trading Platform Application Scenario Replication｜Standardised Product Trading｜ Intellectual Property Transfer AI Project Visualisation Development Platform Project Progress Management｜Project Progress Tracing Process Technical Support Government Connecting Platform Manufacturing Empowerment Government Procurement Projects Successful Case Replication 92% 49% 8% Textile Industry 1 AI New Experience in Fabric Inspection Agriculture 2 Yingde Red Tea Withering Process Practice Animal Husbandry 3 Intelligent "ID Card" for Dairy Cows Construction Industry 4 AI Protecting Tower Crane Safety Urban Management 5 All-weather Road Disease Inspection 1 Textile Industry AI New Experience in Fabric Inspection ❖ Low Efficiency：Manual fabric inspection time is about 15 meters/minute, and repetitive tasks such as marking and recording data take more time. Traditional reports require manual calculation and filling. ❖ Poor Quality：Poor detection of defects, manual inspection is prone to fatigue and subjective errors, with an average detection rate of about 70%. ； ❖ High Cost：High recruitment, management, and training costs. Information recognized by humans cannot be effectively transmitted, and the industry faces pain points such as forming unified standards. ❖ Image Acquisition Module collects image data of fabrics, including color, texture, shape, etc ❖ Image Processing Module preprocesses the collected images, including detail enhancement, noise removal, contrast adjustment, etc. ❖ Intelligent Decision-Making Module comprehensively judges the quality and qualification of fabrics based on factors such as fabric material, color, size, and historical inspection data, and generates inspection reports that meet requirements. This solution is widely applicable to fabric quality inspection at various stages of textile industry manufacturing, printing and dyeing, garment making, etc., and is suitable for inspection of surface defects and color differences of knitted and woven plain fabrics. More accurate, faster inspection efficiency, and lower inspection costs make intelligent fabric inspection the best choice for the future textile industry. 2 Agriculture Yingde Red Tea Withering Process Practice Guangdong Hongyan Tea Industry Co., Ltd. is an important enterprise in the Yingde red tea industry and a technology transformation platform of the Tea Research Institute of Guangdong Academy of Agricultural Sciences. By relying on technological advantages, it is committed to R&D and production, representing the highest level of specialty famous tea products in Guangdong. However, Hongyan Tea also faces industry pain points such as tight labor in tea gardens, inconsistent picking standards for fresh leaves, uneven withering standards for red tea, and contradictions between product standards and tea garden production. Withering is an important process for forming the quality of black tea. Currently, in traditional withering processes, fresh leaves are spread on withering troughs, and production personnel control the air volume and time of the withering trough blower to wither the fresh leaves. Since different tea makers have different judgments on the activity of withering, it directly affects the quality of each batch of tea leaves. By combining infrared thermal imaging technology, AI intelligent recognition models, and high-definition camera monitoring technology to form intelligent withering equipment, it is applied to tea withering. The blower's start and stop state and air volume size are automatically controlled based on the changes of various factors in the withering process of fresh leaves, accurately mastering the withering standards of tea leaves. "Smart Tea Processing" is one of the key research and implementation objects in the future tea industry. Therefore, the establishment and application of intelligent withering equipment have great prospects. Moreover, by establishing models for the standard state of appropriate withering of tea leaves, it can more accurately determine the real-time state of fresh leaf withering. Additionally, the withering equipment can automatically control the blower's start and stop state based on the water content and physical state of the fresh leaves, truly realizing "fully intelligent withering of fresh leaves." 3 Animal Husbandry Intelligent "ID Card" for Dairy Cows The client focuses on movable property pledge supervision business, specializing in movable property pledge sup