The system, “Occluded Resistant Face DetectionRecognition System”, contains different scale detectors for calculating face locations. Additive angular margin loss is added into the training phase for achieving high efficiencyaccuracy. The system can achieves 80 accuracy under the 50 face occluded.

The technology is an integrated solution incorporating the retrievable personality for multimodal emotion sensing for spoken services. This framework provides a real-world flexibility that enables the estimation of the target speaker emotion states without manual personalization.

The dialogue system is the main subsystem of the visually impaired navigation system, which provides destinations for the navigation system through multiple dialogues. We use the knowledge graph as the basis for reasoning. In terms of close-range navigation, deep learning technology is used to develop RGB camera detection depth algorithm, indoor semantic cutting algorithm, integrated detection depth estimationindoor semantic cutting in indoor obstacle avoidance, etc. The whole system uses the CellS software design framework to integrate distributed AIoT systems.

In order to solve the undiscovered problem that caused by the fall ofelders, we develop a deep learning based systemnamed it 「SkyEye」,which includes our own sensor 「AI Falling Image Sensor」, a cloud sever that storespushes information about fall eventsa mobile app that communicates with users.

Combined with monitors in markets, stores,transportation stations, AI image recognition technology is used to detect crowdspeople without masksbroadcast warnings,assist the government in epidemic prevention-related work.

This project uses AIoT technology to face the occurrence issue of incorrect preｓｃｒｉｐｔion drugs of the drug dispensing phase in hospital dispensing Room. The proposed learning-object-detection based deep technology AI mechanism can automatically identifyverify the correctness of the drugs in the drug bag, thereby significantly improving the accuracy of drug dispensing in Hospital.

To resolve public safety and security issues, NCHC developed intelligent video analytics and retrieval system, which can integrate the attribute analysis and recognition of vehicles/pedestrians from CCTV surveillance cameras, utilize TWCC high-speed operation in developments such as AI image semantic analysis and video synopsis technique, to effectively and rapidly search for key targets(vehicles/pedestrians) and provide short clips of the targets’ moving path simultaneously, and as a result fulfill the quick search application service of technological surveillance and Investigation.

The proposed indoor radar sensing system includes two enabling technology, activity recognitionrespiration rate estimation. In the former, the proposed framework consists of four major components: denosing, enhanced voxelization, data augmentation,dual-view machine-learning to lead to accurateefficient human-activity recognition. In the latter, the proposed system leverages the variation of the phase information of a specific frequency bin of the range profiles,proposes a dynamic adaptive respiration waveform filtering algorithm to improve accuracy.

An automated agricultural machinean intelligent crop production decision-making system are developed for greenhouse asparagus growth monitoring. Automationenergy-saving machinery technologiesInternet of Things/information communication techniques are adopted to achieve the goal of "saving laboreffort"improve "management technologies". The developed machinesystem can greatly improve the qualityquantity of domestic asparagus production,achieve the dual goals of stabilizing the outputenhancing quality of harvested asparagus.

This technology integrates 1000+ times of UAV imaging experiences with labeled rice lodging images for training. A rice lodging recognition model using deep learning reaches 90 accuracy. The recognition model can be deployed in a microcomputer mounted on UAVs to implement edge computing. While taking aerial images, the inference can be completedreveal lodging areadamage level in-time.

The Air/Ground cooperation for optimal rice harvesting model is established to provide a visual harvesting decision service on a cloud platform. Drones and mobile devices are employing to estimate grain moisture and forecast the variation of harvest moisture content (HMC) in the coming days by huge amounts of imagery data, deep learning algorithms, and weather forecasts. This model can benefit in several aspects, such as setting an accurate and comprehensive optimal harvest schedule, reducing the cost of agricultural apparatus and barn ovens, ensuring the rice quality, and maximizing farmers' benefits. The potential value of the model practice could be more than a billion in Taiwan.

In the face of the COVID-19 pandemic, many restaurants have been closed for maintaining anti-epidemic measures. NCHC combined multiple AI technologiesdeveloped an automated service experience to provide a way for people to still enjoy the services received at restaurants while maintaining social distancing. To achieve this, we＇ve incorporated technologies including the following: NLP in AI, Cloud-Based Smart Point Cloud Processing (CSPCP), method for face searching in images, intelligent video analysisretrieval,artificial intelligence of things (AIoT).

The pills detection system is built based on the Feature Pyramid NetworkConvolution Normal Network (CNN). Two-stage CNN architecture is used for pills localizationpill classification. The accuracyexecution time of the pill detection are 900.02s, respectively.

This system uses the innovative AIoT application to target annoying scalp maintenance problemsdevelops an intelligent scalp detectionmanagement system. The core of the technology is to use deep learning-based object detection models. Tens of thousands of microscopic image data sets that are labeledtrained for scalp symptoms. As a result, the scalp image recognition module is develop, such as dandruff, hair loss, oil, and inflammation, etc. Hence this system can provide scalp detection, and maintenance effectiveness tracking functions lead scalp maintenance services to a new level of intelligent management.

"1. Automatically generate Verilog code tools based on the hardware architecture of convolutional neural networks: 4 different hardware architectures (output stationary, weight stationary, Tree architecture, NVDLA) can be generated for the currently more commonly used DNN networks. 2. Visual performance index analysis tool: According to the selected DNN model,the choice of hardware architecture specifications, analyze the performance index."

Super fast terapixel server allows multiple users to viewanalyze gigapixelterapixel images with mobile devices, such as iPhones, iPads, Android mobiles,laptops at the same time. Users could annotateanalyze the data using built-in tools such as 3d effect, color deconvolution, color quantification, nuclei/object detection, lengtharea measurement,quantitative reporting system.

(1) Image recognition3D positioning of the pest: Modularize the system for testing on the orchard. (2) High efficiency laser pest control scanning system: It was developed to be suitable for windlessbreezy environments. (3) Enhance the economic tracked vehicle for hillside field: Self-developed power designsteering control based on dynamic model,(4) Field UGV control platform: Based on the preciseaccurate positioning in field, when robot operates on the reference path which planned by map, it can do path trackingdynamically obstacle avoiding in real-time.

During the pandemic, it is necessary to reduce contactwear masks, but identification is still required. The original technology includes the development of inertial measurement units beneath the walking surface, real-time detection of gait,accurate predictions of gait parameters with machine learning models. Most importantly, our technology can use a small number of samples (only one footstep data) for identifications. Non-contact, non-monitoring,rapid identification are going to be the future applications in smart buildings/homes.

”AI2 Robot City” is a game-based learning kit for primary and secondary school students, which combines AI image-recognition teaching tool of MIT App Inventor and the computational thinking board game named as "Robot City." Through this learning kit, users will learn to make smart cars, create image-recognition models, write and perform mobile application. Users will learn to write a program to recognize the personal cards in the board game, and furthermore to control the smart cars with blue-tooth and compete in the computational thinking board game.

1. CNN model architecture optimization techniques: Propose an Agilev3-L architecture, with a performance index of mAP@50:95FPS:FP16 up to 171.79 2. GOP-mode acceleration scheme for real time inference: employ GOP-modeTRACKING algorithms to enhance the processing rate. 3. Propose a rapid data labeling system HiTag.

The technology aims to develop an AI-based integrated system for emotional detections (anger, sadness, happiness,neutral)multimodal physiological signals (ECG, EEG,PPG),apply to patients with cardiovascular disease. Patients monitor their emotionalphysical statusadminister bio-neuro-feedback to improve their well-being, track disease progression,prevent adverse prognosis.

This study built an automatic insect pest image identification system based on tiny Yolov3 deep learning model. By optimizing the tiny Yolov3 detection model, images of insect pests on scanned sticky paper can be automatically identified. The system achieves a testing accuracy of 0.93, 0.90 for whitefliesthrips respectively.

In this study, the embedded LiDARAI recognition of smart headlight for autonomous driving is demonstrated. At driving process, the vehicle can provide artificial intelligence to identify the type, distancespeed of the object to be measured based on the fusion data of the LiDAR point cloudoptical image of the environment,then feedback to the headlight control system. Finally, this smart headlight would perform the suitable lighting mode during the daynightfit the automotive lamp safety regulationsstandards (ECE R112SAE J3069).

This project integrated with industrial foresight technologies, including UAV, artificial intelligenceimage recognition, to collect real-time images, apply algorithm in evaluation, link the technology of IOT (Internet of Things)environment sensing,use unmanned vehicle to conduct controlling work.

Through patented wearable sensor fusion technology, human 3D motions can be captured in real-time which allow 7-DOF robotic armdexterous robotic hand to perform anthropomorphic motiongrasping action. Integrated with eye-in-hand visual module, the adaptive gripper can intelligently decide optimal grasping strategy to ensure success grasping of arbitrary shape of objects in smart manufacturing applications.

Deep-Learning-Based Pupil Center DetectionTracking Technology for Visible-Light Wearable Gaze Tracking Devices

The Artificial Intelligence Underwater MonitoringManagement system is a system that integrates aquacultureinformation technology, which can collect image data of shrimp 24/7 by using IoT devices, then next enhance the images that collected under different light, water qualityturbidity. The system will recordanalyze the number, size, healtheating status of white shrimp, checking the amount of leftovers feed on the feeding tray, then manage the feeding strategycontrol water quality based on the above analysis.

Low-power deep learning accelerator integrates neural network design / training, model compressionaccelerator design. It uses a small amount of computationmemory footprint to realize high performance on the edge device. The image semantic segmentation technology can reach 80 frames per second (resolution:1024*2048) to meet real-time requirements for autonomous driving.

This invention presents a "next-generation augmented reality navigation system” using the generative adversarial network-long short term memory network (GAN-LSTM) framework with integrated GPS module to implement a novel AR navigation system. Unlike the presented AR navigation system, the virtual guided path is "autonomously generated" in captured image rather than superimpose on the image by using the pre-rendered 3D content, which not only provide a more authenticcorrect AR effect to user but also earlycorrectly guide the driver when driving in complex road traffic environment.

The developed Unmanned autonomous underwater vehicles (AUV) included 3-D modeling technology, wireless power transfer between sub-system in AUV, GPS underwater locate system, low complexity frequency domain on ocean floor analyze with AI network, underwater objects detection network, image dehaze network on groundunder water, underwater color correction network, underwater objects classification network, underwater optical characteristic algorithmLED color compensation system.