Smart Manufacturing

Industry 4.0 accelerated the development of the manufacturing industry; various parts of the processes are integrated with technologies including cloud, Internet of Things (IoT), sensors and big data analysis, which made manufacturing smarter. Leadtek AI team has rich experiences that can help your system learn how to collect big data, analyze the information and actively learn from experience continuously, and further learn how to predict and avoid possible errors.

According to statistical data, over 90% recognition rate was achieved when actually tested on production lines. The turnover of departments increased by an average of approximately 3 times within a year the system was implemented.

AI implemented processes：

• Flaw detection analysis
• Fault warning analysis
• Quality control
• Process quality control
• and more

Smart Medical & Healthcare

For MRI , CT, X-ray, OCT and other types of imaging, use medical image analysis with Leadtek AI deep learning solution to enhance image analysis performances, allowing pathologists or physicians to find out locations or areas where lesions might occur, and provide even more precise medical treatment services.

By using the parallel computing features of the GPU, the algorithm calculation time can be shortened, which is useful to perform high performance DNA alignment, drug synthesis screening and more.

AI implemented processes：

• Precision Medicine
• Medical image analysis
• Skin image analysis
• Clinical data analysis
• Pathological Examination
• Biological Data Analysis
• High performance DNA alignment
• Drug Analysis
• Personalized Medicine
• and more

Smart Agriculture

Use integration of identification application software and hardware, and customized integration of deep learning software, sensors and calculation hardware needed for the field based on the site and customer needs to achieve identification of common pests and weeds or detect pesticide residues; effectively using big data to reduce unnecessary pesticide spraying and create friendly environment agriculture.

AI implemented processes

• Pest identification
• Weed identification
• Pesticide residue detection
• and more

How GPU speeds up AI &
deep learning computation

cuDNN

NVIDIA cuDNN is a GPU accelerated library of primitives for deep neural networks that provides highly tuned implementations for standard routines; these routines are commonly used in DNN applications, including Convolution, Pooling, Softmax, Sigmoid, ReLU or TANH. It is supported by most mainstream deep learning frameworks nowadays, such as Caffe, TensorFlow, MxNet, CNTK and PyTorch.

NVLINK

GPU can quickly process large amounts of data, but the bottleneck for performance usually occurs at the bandwidth of the PCI-E transmission interface. NVLINK is a high-bandwidth interconnection that enables ultra-fast communication between the CPU and GPU, and between GPUs. The technology allows data sharing at rates 5 to 10 times faster than traditional PCIe Gen3 interconnections.

NCCL

The systems of data centers usually have more than one set of GPU cards. In order to exert maximum performance through the GPU, the use of NCCL technology can drive multi-GPU to perform computations simultaneously, reducing the AI development time significantly.

TensorRT

High performance inference engine that is used specifically for the deep learning deployment stage. It converts models into INT8 or FP16 in order to perform faster inference, allowing developers to focus on the development of AI models and not the adjustment of computation performance.