How does Intel AI help repair the ruins of the Great Wall, the mystery behind it found
Rame f rameborder="0" width="640" height="498" src="https://v.qq.com/if rame/player.html?vid=p0718e17ovv&tiny=0&auto=0"> rame>
When artificial intelligence reaches the crossroads of technology and humanity, what kind of energy does it emit? You will find that there is no period in the past that will be closer to us than the current artificial intelligence, because it is solving the real problems in life in a near-revolutionary way.
Now, there is a thorny question in front of the China Cultural Relics Protection Foundation: How to repair a long-term, long-distance, already a broken wall of the Great Wall?
If it is in accordance with the traditional method, whether it is surveying, measuring, or post-processing, it can only be done by manpower with a ruler and a total station. It takes time and effort, especially the Great Wall of the Jiankou is mostly located on the cliff of the dangerous peak, and the surrounding grass is dense. It is dangerous to have people want to reach the construction site. In addition, traditional manual measurements do not reflect the full picture and details of the Great Wall, which is not conducive to the protection unit to make more accurate maintenance programs.
Nowadays, the project “Artificial Intelligence Helps Repair the Arrows Great Wall†has been carried out in cooperation with Intel China Research Institute, Intel Data Center Business Unit and Wuhan University Key Laboratory of Surveying and Mapping. The new method can digitally restore the state before and after the repair of the Great Wall, which is conducive to repair protection and further research. With the participation of Intel's artificial intelligence technology, the protection of the Great Wall and more Great Walls will reach new heights.
How to use Intel AI to help the Great Wall repair, mainly in three steps:
1, collecting high-precision images
In the Arrows Great Wall Protection Project, Intel's latest Falcon 8+ drones were used to perform overall and partial aerial photography and accurate imaging of the Great Wall.
2, 3D modeling and artificial intelligence recognition of damaged parts
Quickly analyze and process high-resolution image data with the latest Intel Xeon server to produce a complete high-precision Great Wall image 3D model, using artificial intelligence algorithms to identify the parts that need to be repaired on the 3D model, and provide damage such as cracks and landslides. The measurement data is used to guide physical repairs.
3, 3D model of artificial intelligence digital repair
Based on the 3D model damage identification, the latest 3D model is used to combat the generation network, and the regression convolution network is used to digitally repair the defect parts of the city wall, and provide guidance and reference data for the actual repair and maintenance of the Great Wall.
This will be a brand new exploration, advanced drone aerial photography and artificial intelligence technology involved in the repair and protection of heritage buildings, Intel's computing technology is deeply involved. The data shows that only 700 meters of the Great Wall, the Falcon 8+ UAV collected tens of thousands of high-resolution images, the original data is more than 200GB, the entire process will access these data frequently, and will generate more than 100GB of intermediate and simulation Data, even for high-performance computing, is extremely complex to handle such a large amount of data.
The solution also involves a variety of AI algorithms, including visual feature extraction and indexing, camera parameter recovery, beam adjustment, dense matching, geometric model mesh generation, deep neural network 2D and 3D model training, texture synthesis, and more.
Intel's solution is based on the Xeon Xeon scalable processor, Intel SSD, combined with OpenMP/MPI parallel optimization technology, Intel Core Deep Neural Network Math Kernel Library (MKL-DNN) optimized for Intel CPU, and Tools such as Tensorflow, an Intel architecture-optimized deep learning framework, efficiently implement Great Wall 3D modeling and digital restoration, and achieve centimeter-level accuracy.
In the process of 3D modeling and digital restoration of the Great Wall, large-scale equation iterative calculations are needed. Some equations based on large-scale sparse matrices have convergence stability problems. At this time, the role of the large-scale matrix computing library MKL is highlighted, which not only can improve the computational efficiency, but also greatly improve the stability of complex computing.
Today, the MKL-DNN library developed by Intel has been widely used in popular deep learning frameworks such as Tensorflow and Caffe. It can be said that for the solutions implemented by different algorithms in the deep learning field, the Intel Xeon architecture is an ideal choice for supporting such a variety of algorithms in a comprehensive, efficient and low-cost manner, and can significantly improve the efficiency and speed of the artificial intelligence repairing the Great Wall.
Intel Xeon Server provides a full suite of development tools for AI developers, allowing developers to configure memory requirements on demand based on deep learning data complexity. On this basis, Intel (China) Research Institute and Wuhan University will develop deep learning algorithms for Great Wall defect/crack identification and location, digital restoration, including:
1. Great Wall Defect/Crack Identification and Location
For damage and crack types, the researchers collected and calibrated the normal and damaged Great Wall 3D models and obtained enough sample data to train the deep learning network. The network uses a combination of regression and convolution. Aiming at 2D views and profiles of different perspectives of 3D models, a large number of data samples are trained and analyzed to form a recognition ability for typical damage modes.
2, Great Wall digital model virtual repair
When a damaged part of the Great Wall is identified, the AI ​​will perform a digital virtual repair, generate 3D repair effects and brick wall texture on the damaged model, and obtain the engineering quantity data required for physical repair as a physical repair. Reference suggestions.
In digital restoration, a large amount of 2D/3D model generation technology will be applied. Whether it's 2D or 3D model generation network training, its data input and computational complexity are amazing, and only Intel Xeon servers can provide complete support. At the same time, the digital repair of Intel AI will follow the "repair of old and old" cultural relics repair principle, provide detailed location, effect and estimation of required engineering quantity for the Great Wall repair project, as an effective reference and comparison of actual engineering.
All in all, in the process of helping repair the Great Wall, Intel provides a complete manual from front-end data collection to back-end deep network training and solution generation, integrating measurement tools, artificial intelligence algorithms, and high-performance computing platforms. Intelligent heritage repair and protection solutions.
In fact, not only repairing the Great Wall, other artificial intelligence applications, Intel's AI solution is also a good choice. Because whether you are a data scientist or an IT architect, you can develop simple and efficient on-demand development on your familiar CPU platform, making AI development and application simpler and more practical.
It is precisely this high-performance general-purpose AI computing platform, and its practical application to engineering and product capabilities, will help more AI solution providers, bold breakthroughs in the field of AI applications, launch an AI application Great times.
Artificial intelligence, starting from Xeon. The repair and protection of cultural relics is such a vivid example, and it is only a beginning.
Switching Power Supply, CCTV Switching Power Supply, LED Power Supply,CCTV Camera Power Supply
Chinasky Electronics Co., Ltd. , https://www.cctv-products.com