Facial Landmark Detection Using Local Regions and Joint Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial key point detection methods that use the entire face image as a detection object are inefficient, leading to high time consumption and frame freezing during applications like makeup applications.
Innovation Solution
The method involves determining partial images containing key points from a face image, using trained neural networks to identify candidate points, and jointly constraining these points to improve detection efficiency and accuracy, employing a linear model to avoid complex rotation operations and ensure global optimal solutions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire face image is used as the detection object, then the detection coverage is complete, but the detection efficiency is low and time consumption is high
Solution Approach 1:
The patent divides the face image into multiple local regions (eye region, eyebrow region, nose region, mouth region, cheek region) and performs key point detection separately on each region. This segmentation approach reduces the computational complexity compared to processing the entire face image, while still ensuring complete detection coverage across all facial key points.
2Measurement precision
If the entire face image is processed, then all key points can be detected, but the calculation amount is large and time consumption increases
Solution Approach 1:
By segmenting the face into five distinct regions and processing each region independently with region-specific neural networks, the patent significantly reduces the calculation amount. Each local region requires fewer computational resources than processing the entire face image, thereby reducing overall time consumption while maintaining complete key point detection.
Solution Approach 2:
The patent applies partial action by focusing computational resources only on relevant local regions rather than processing the entire face image. Each region is processed with appropriate depth and complexity, avoiding unnecessary computation in areas where key points are not present or are less critical.
3Measurement precision
If traditional detection methods are used, then comprehensive key point detection is achieved, but frame freezing occurs during makeup applications
Solution Approach 1:
The patent segments the face processing into five parallel local region detections that can be executed concurrently or in optimized sequence. This segmentation enables better resource utilization and reduces frame processing time, preventing frame freezing during makeup applications while maintaining comprehensive key point detection accuracy.
4Measurement precision
If global facial features are used for detection, then overall face structure is captured, but detection efficiency is reduced
Solution Approach 1:
Instead of using global facial features that require processing the entire face image, the patent segments the detection into five local regions. Each region is processed by a specialized neural network that captures the essential structural characteristics of that specific area, achieving both efficient processing and accurate structural understanding.
Solution Approach 2:
The patent applies local quality by using different neural network configurations optimized for specific facial regions. Each region (eye, eyebrow, nose, mouth, cheek) has its own detection model tailored to its characteristics, improving detection efficiency by avoiding the overhead of global feature extraction while maintaining accurate facial structure understanding.
Data Source
Figure 1~2
Figure 3~5
Figure 6
AI summary
The present application relates to a facial landmark detection method and apparatus, a computer device and a storage medium. The method comprises: acquiring a facial image to be detected, wherein said facial image is a facial image of a frame to be detected; determining, according to said facial image, local images, respectively comprising each landmark, in said facial image; respectively determining, based on each local image, a candidate point of the corresponding landmark in each local image; and performing combined constraint on the candidate point of each landmark, and determining each facial landmark. For local images, respectively comprising each landmark, in an entire facial image to be detected, candidate points of corresponding landmarks in the local images are respectively determined. Therefore, the amount of calculations can be reduced, and the efficiency of determining the candidate points of the landmarks can be improved. Therefore, the efficiency of detecting each facial landmark can be improved.