When data is sufficient, Deep Learning can be used to generate models for certain scenarios and applications.
Deep Learning with Small Data
When data is insufficient, migration can be used to generate a relatively general model using data from other similar fields.
Deep Learning with Noisy Data
When the labels of data samples have errors, recursive training and label correction can automatically restore those noisy sample labels.
Semi-Supervised Deep Learning
When some labels of data samples are missing, semi-supervised deep learning can gradually estimate labels for the unlabeled data.
Deep Learning with Zero Data
When no obvious data can be found in the field, Generative Adversarial Network is able to hallucinate data and solve such problems in a certain degree.
Face Recognition
Face Detection
Supports all-posture real-time face detection.
Facial Landmark Positioning
Can precisely find out the position of more than 80 facial landmarks in the image such as eyes, noses, eyebrows and mouths in real time.
ID Verification
Can verify whether the ID photo matches the person at site.
Dynamic Face Detection
Can verify whether the people in the surveillance video is in the registered list
Large-Scale Face Searching
Can precisely and rapidly find out the best match in the registered library of hundreds of millions of faces.
Liveness Verification
Can automatically judge whether the face in the video belongs to a live person; prevents fake ID attacks.
Face Attribute Analysis
Can distinguish the gender, age, emotion, race, and accessories of the captured person
UAV Vision
Ground Target Detection and Recognition
Ground targets can be detected and recognized through high-resolution aerial cameras on UAVs.
Vision-Based Distance Measurement
UAV vision module can evaluate its distance to the captured objects, and avoid obstacles.
Robust Multi-Object Tracking
UAV can calculate movement parameters of multiple objects in the video image sequence, and effectively understand the movement of those objects.
Fine-Grain Image Searching
Given an aerial photo of the target, UAV vision module can locate the target via aerial camera.
Ground Target Detection and Recognition
Ground targets can be detected and recognized through high-resolution aerial cameras on UAVs.
Vision-Based Distance Measurement
UAV vision module can evaluate its distance to the captured objects, and avoid obstacles.
Robust Multi-Object Tracking
UAV can calculate movement parameters of multiple objects in the video image sequence, and effectively understand the movement of those objects.
Fine-Grain Image Searching
Given an aerial photo of the target, UAV vision module can locate the target via aerial camera.
Video Structural Analysis
Object Detection and Re-IdentificationCan detect, track, and precisely re-ID the target through video content structuring technology.
Object Attribute AnalysisCan analyze detailed attributes of the target using video content structuring technology.
Object Trajectory AnalysisCan analyze the trajectory of the target through video content structuring technology.
