Shared Memory Neural Networks for Multi-Object Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking methods in traffic situations face challenges in accurately tracking multiple objects, especially when objects are partially or completely concealed, and their trajectories overlap, leading to a loss of overall context and reduced accuracy.
Innovation Solution
The method employs a shared differentiable neural computer (DNC) where each object is assigned a neural network that uses a shared memory to track and characterize objects, allowing for precise identification and behavior analysis, including gestures and facial expressions, while maintaining context through seamless information exchange between networks during training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single neural network is used to track multiple objects, then device complexity is reduced, but tracking accuracy deteriorates when objects are concealed or trajectories overlap
Solution Approach 1:
The system divides the tracking task by assigning a dedicated neural network to each object, creating multiple specialized trackers instead of one general tracker. Each neural network focuses on tracking a specific object, which improves accuracy when objects are concealed or have overlapping trajectories, as each network maintains independent object context and history.
2Measurement precision
If multiple neural networks are assigned to track each object individually, then tracking accuracy is improved, but device complexity increases
Solution Approach 1:
The shared memory structure serves multiple neural networks simultaneously, providing a universal resource that enables each network to access object history and context without requiring separate memory systems. This multi-functional shared memory reduces the overall system complexity despite having multiple neural networks, as it eliminates the need for duplicate memory structures for each network.
3Reliability
If individual neural networks track objects independently, then tracking robustness to concealment is improved, but loss of overall context information occurs
Solution Approach 1:
The shared memory acts as an intermediary between individual neural networks, allowing them to exchange information about object contexts, trajectories, and behaviors. This mediator enables each network to maintain its independent tracking robustness while simultaneously accessing overall scene context through the shared memory, thus preventing information loss.
4Loss of information
If shared memory is implemented for information exchange between neural networks, then overall context retention is improved, but computational overhead increases
Solution Approach 1:
The system merges the memory structures of multiple neural networks into a single shared memory system. This consolidation allows all neural networks to access the same memory locations, enabling efficient information exchange about object contexts and trajectories. The merging reduces redundant computations and memory operations compared to having separate memory structures for each network, thus lowering computational overhead while retaining overall context.
Data Source
AI summary
A method for tracking and/or characterizing multiple objects in a sequence of images. The method includes: assigning a neural network to each object to be tracked; providing a memory shared by all neural networks, and designed to map an address vector of address components, via differentiable operations, onto one or multiple memory locations, and to read data from these memory locations or write data into these memory locations; supplying images from the sequence, and/or details of these images, to each neural network; during the processing of each image and/or image detail by one of the neural networks, generating an address vector from at least one processing product of this neural network; based on this address vector, writing at least one further processing product of the neural network into the shared memory, and/or reading out data from this shared memory and further processing the data by the neural network.