Neural Network Face Detection With Tiled Feature Anchors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face detection algorithms using neural networks face challenges with high computational requirements and power consumption, making them unsuitable for many in-device applications.
Innovation Solution
A neural network is trained using a three-phase data set approach, including dividing images into tiles, generating feature anchors, detecting and labeling target features, and refining the network with a compressed convolutional architecture optimized for efficient processing on devices without dedicated neural processing units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional neural networks are used for face detection, then detection accuracy is improved, but power consumption and computational requirements increase
Solution Approach 1:
The image is divided into multiple tiles, and each tile is processed independently by the neural network. This segmentation allows the system to process only relevant portions of the image at high resolution, reducing overall computational load and power consumption while maintaining detection accuracy for faces within each tile.
Solution Approach 2:
The system performs full neural network processing only on selected tiles that contain potential face regions, rather than processing the entire image. This partial action approach reduces the total number of computations required, lowering power consumption while maintaining accurate detection where needed.
2Measurement precision
If traditional neural networks are used for face detection, then detection accuracy is improved, but device complexity increases
Solution Approach 1:
By dividing the image into tiles and processing only selected tiles with potential face content, the system reduces the total computational complexity. The neural network is applied to fewer, smaller regions rather than the entire image, making the device more manageable while preserving detection accuracy.
Solution Approach 2:
The system performs preliminary processing to identify candidate regions containing faces before applying the full neural network. This preliminary action filters out irrelevant areas, reducing the computational complexity of the main detection process while maintaining high detection accuracy for actual face regions.
3Reliability
If comprehensive data sets are used for training, then detection robustness is improved, but training time increases
Solution Approach 1:
The training process uses segmented tiles from images, allowing the system to efficiently process and learn from multiple regions across diverse images. This tile-based approach enables comprehensive training data utilization without linearly increasing training time, as the same neural network processes multiple tiles in batch operations.
Data Source
AI summary
A method of training a neural network for detecting target features in images is described. The neural network is trained using a first data set that includes labeled images, where at least some of the labeled images having subjects with labeled features, including: dividing each of the labeled images of the first data set into a respective plurality of tiles, and generating, for each of the plurality of tiles, a plurality of feature anchors that indicate target features within the corresponding tile. Target features that correspond to the plurality of feature anchors are detected in a second data set of unlabeled images. Images of the second data set having target features that were not detected are labeled. A third data set that includes the first data set and the labeled images of the second data set is generated. The neural network is trained using the third data set.