Target recognition system and method based on time slice convolutional neural network
A convolutional neural network and time slicing technology, applied in biological neural network models, neural architectures, character and pattern recognition, etc., to improve recognition efficiency and accuracy
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Example Embodiment
[0049] Example 1:
[0050] Based on time-slicing target convolutional neural network recognition systems, such as figure 1 As shown, the system includes:
[0051] Dividing module 101 event stream, the event stream samples for slicing, forming a set of events, the event by the event collection is represented in a pseudo image as the dummy image in a time slice convolutional neural network input channel to the first splice feature FIG re-assigned a different weight for each channel to obtain a second characteristic diagram;
[0052] Feature extraction module 102, a second feature of an input to the convolutional neural network in the time slice feature extraction becomes the predetermined specification is characterized in FIG;
[0053] Classification module 103, wherein the predetermined specification is for converting a vector in FIG obtain maximum probability category, the recognition result as a target.
[0054] Specifically, the stream event be segmented to form a set of events, c...
Example Embodiment
[0072] Example 2:
[0073] An object identification system time slicing based on convolutional neural network, such as still figure 1 Shown, the system comprising: a segmentation module 101 an event stream, the event stream samples for slicing, forming a set of events, the event by the event collection is represented in a pseudo image as the dummy image slices convolutional neural network in a time splicing a first input channel characteristic diagram, again for each channel assigned different weights, to obtain a second characteristic diagram; feature extraction module 102, a second feature of an input to the convolutional neural network in the time slice feature extraction FIG characteristic becomes the predetermined specification; classification module 103, configured to convert the predetermined specification is characterized FIG vector, the probability of obtaining the largest category, the recognition result as a target.
[0074] Preferably, the slicing event stream during e...
Example Embodiment
[0082] Example 3:
[0083] Based on time-slicing convolutional neural network object recognition method, such as image 3 As shown, including the following steps:
[0084] Sl, the dicing step event stream, the event stream samples sliced, forming a set of events, the event represented by a set of event representation to pseudo image, the pseudo-image in a time slice convolutional neural network input channel to splice FIG first feature, re-assigned a different weight for each channel to obtain a second characteristic diagram;
[0085] S2, feature extraction step, wherein the second time slice of an input to the convolutional neural network feature extraction becomes the predetermined specification is characterized in FIG;
[0086] S3, the classification step, wherein the predetermined specification is converted FIG vector, the probability of obtaining the largest category, the recognition result as a target.
[0087] As an embodiment of the convertible, based on the time slicing co...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap