Object Detection solution

AIPro's object detection solution ensures very high detection accuracy by intensively learning the target object required by the clients. It provides a customized deep learning object detection solution by reflecting the type of object to be detected, accuracy, and operation environment, and can simultaneously detect the type of object and the position of the object in the image. Our solution can be applied to any hardware of the clients since it can respond to environments such as servers, desktops, mobiles, and embedded devices. With our customized solutions, the clients can achieve safer Advanced Driver Assistance Systems (ADAS), more accurate CCTV remote monitoring, and more efficient factory automation.

System environment Server, desktop, mobile, embedded device
Supported OS Windows. Linux. Android. iOS
Detection accuracy 99% ~ 90% (varies by models)
Reference time 0.025 s/frame ~ 0.11 s/frame
Detectable objects Model 1: 80 common objects (People, cars, dogs, traffic lights, bags, etc.)

Model 2: 13 main objects (People, cars, buses, taxis, motorcycles, bicycles, etc.)

Model 3: 6 human body parts (Body, head, face, hand, mouth, foot)

Other customized model production for the clients
Number of simultaneous detected objects Up to 100
Application examples CCTV control station, driver assistance system, factory automation, etc.

Object tracking solution

AIPro's object tracking solution enables real time tracking with resource-limited devices, even if the objects in the scene are moving rapidly. Our object tracking solution shows a high tracking success rate by calibrating the camera movement even when shooting a target in a mobile device environment (Android, iOS). Our object tracking solution makes it possible to implement a more stable system by preventing the waste of excessive hardware resources up to 10 times due to frequent object detection operations. In addition, by using object tracking technology, services such as counting pedestrians on the street, counting the number of visitors to the building, monitoring the flow of vehicles on the road, and monitoring the movement of objects of interest can be provided.

System environment Server, desktop, mobile, embedded device
Supported OS Windows. Linux. Android. iOS
Resolution All resolutions supported by the device
Tracking time 0.005 s/frame ~ 0.01 s/frame
Types of objects that can be tracked All detected objects
Number of objects that can be tracked simultaneously No limit
Application examples Counting pedestrians on the street, counting the number of visitors, monitoring vehicle flows on the road, and monitoring the movement of objects of interest

Super resolution solution

It is a technology that converts low-resolution low-quality (blurred images) images into high-resolution high-quality images using deep learning CNN technology. AIPro's super resolution (SRCNN) technology supports resolution conversion from a minimum of 4x to a maximum of 64x magnification. We highly train the state-of-the-art deep learning models on a vast data set to support optimal high-resolution transformations. In addition, we minimized the high-resolution conversion delay of the trained SRCNN model using the latest deep learning optimization technologies. Clients can use this solution to provide user services that enable users to observe and read ultra-high-definition images that go beyond the limits of physical observation.

System environment Server, desktop, mobile, embedded device
Supported OS Windows. Linux. Android. iOS
Resolution All resolutions supported by the device
Conversion time 0.03 s/frame ~ 1.00 s/frame
Supported inputs Image, video
Application examples Digital zoom, microscope, high definition image conversion, image & video streaming

Image Super Resolution (x4)

click the image

superresolution_x4_img1 superresolution_x4_img2 superresolution_x4_img3 superresolution_x4_img4

Video Super Resolution (x4)

ROI-SR

AIPro is the only company in Korea developing hybrid ROI-SR technology, which is the fusion of Object Detection and Super Resolution technologies. AIPro ROI-SR 1) automatically detects the target object from the input image, 2) and selectively enhances the resolution of the detected object region. By using this solution, it is possible to supplement the visibility of a relatively small target object, and to dramatically increase the recognition accuracy, even with limited resources.v

before_roi now_roi

Face Recognition

Face Recognition (FR) solution identifies personal identity of people in the video by comparing the detected face with registered face images. This solution can be used in various human beings-related fields, such as security and services in airports, hospitals, and banks. Further, our FR solution provide customers with more convenience, higher accuracy and stability in comparison with other biometric systems.

PAR(Person Attribute Recognition)

AIPro has studied the state-of-the-art PAR technologies, and created its own solution. Our PAR solution shows excellent performance in terms of accuracy, speed, and the number of simultaneous detections, since it is a technology developed based on our superior Object Detection technology. This solution supports the recognition of various features, such as gender, age, hair length, clothing, and belongings. Further, the fusion of PAR and our Face Recognition solution enables more services to extract large amount of human information.