Deep Neural Network Object Detection via Dual Mask Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection methods in images are inefficient, particularly in handling overlapping objects and requiring manual design of models, which limits their ability to detect a wide variety of objects with high accuracy and speed.
Innovation Solution
The use of deep neural networks to generate full and partial object masks, allowing for the generation of bounding boxes that are refined based on overlap scores, enabling robust object detection without manual model design and supporting high-resolution, non-rectangular detections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual model design is used for object detection, then the system can be simpler to implement, but the ability to detect a wide variety of objects with high accuracy is limited
Solution Approach 1:
The system uses deep neural networks that automatically learn and design detection models from training data, eliminating the need for manual feature engineering. The network self-adjusts its parameters and architecture through backpropagation, enabling it to adapt to various object types without manual intervention while maintaining high detection accuracy
Solution Approach 2:
The patent employs deep neural networks with multiple layers that automatically adjust their internal parameters (weights and biases) through training. This allows the system to transform from a static manual model to a dynamic adaptive model that can detect diverse object types by changing its learned parameters based on training data
2Device complexity
If traditional object detection methods are used, then the implementation can be simpler, but the speed and accuracy of detection are reduced
Solution Approach 1:
The patent divides the object detection task into multiple stages using a cascaded deep neural network architecture. Each stage processes and refines detection results progressively, allowing the system to achieve high speed and accuracy by breaking down the complex detection problem into manageable segments rather than using a single simple method
Solution Approach 2:
The system transitions from traditional 2D image processing to multi-dimensional feature space by employing deep neural networks with multiple layers. This dimensional transformation enables the system to capture complex patterns and relationships in the data, significantly improving detection speed and accuracy beyond what simpler methods can achieve
3Measurement precision
If deep neural networks with multiple layers are used, then detection accuracy is improved, but the computational complexity and training time increase
Solution Approach 1:
The patent implements backpropagation algorithms that provide feedback mechanisms for training deep neural networks. The network computes gradients of the loss function with respect to its parameters and adjusts weights through multiple layers, enabling accurate detection despite the increased complexity by systematically optimizing parameters through feedback from training errors
Solution Approach 2:
The system performs preliminary training of deep neural networks on large datasets before deployment. This preliminary action pre-adjusts the network parameters and structures, reducing the computational burden during actual detection while maintaining high accuracy. The complex network is prepared in advance through extensive training
4Loss of information
If full object masks are generated for all objects, then complete detection information is obtained, but the processing time and computational resources increase
Solution Approach 1:
The patent generates full object masks only for detected objects rather than processing the entire image uniformly. This partial action approach focuses computational resources only on regions containing objects, maintaining complete detection information for relevant areas while significantly reducing overall processing time and computational expenditure
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting objects in images. One of the methods includes receiving an input image. A full object mask is generated by providing the input image to a first deep neural network object detector that produces a full object mask for an object of a particular object type depicted in the input image. A partial object mask is generated by providing the input image to a second deep neural network object detector that produces a partial object mask for a portion of the object of the particular object type depicted in the input image. A bounding box is determined for the object in the image using the full object mask and the partial object mask.