System and method for object detection

Patent metadata

Field	Value
Publication number	US-11663243-B2
Application number	US-202117161541-A
Country	US
Kind code	B2
Filing date	Jan 28, 2021
Priority date	Jan 28, 2021
Publication date	May 30, 2023
Grant date	May 30, 2023

Abstract

An information handling system for managing detection of objects includes a storage and a processor. The storage is for storing an encoder; a critical class classifier; a general classifier; and a decoder. The processor obtains data that may include one or more of the objects; encodes the data using the encoder to obtain encoded data; obtains a critical class classification for the encoded data using the critical class classifier; obtains a general classification for the encoded data using the general classifier; conditions the encoded data to obtain conditioned encoded data; decodes the conditioned encoded data using the decoder to obtain reconstructed data; makes a determination that the reconstructed data and the critical class classification indicate that the data is an unknown classification; classifies the data as being an unknown classification based on the determination; and performs an action set based on the unknown classification of the data.

First claim

What is claimed is:

1. An information handling system for managing detection of objects, comprising:
storage for storing:
an encoder;
a critical class classifier;
a general classifier; and
a decoder;

a processor programmed to:
obtain data that may comprise one or more of the objects;
encode the data using the encoder to obtain encoded data;
obtain a critical class classification for the encoded data using the critical class classifier;
obtain a general classification for the encoded data using the general classifier;
condition the encoded data to obtain conditioned encoded data;
decode the conditioned encoded data using the decoder to obtain reconstructed data;
make a determination that the reconstructed data and the critical class classification indicate that the data is an unknown classification;
classify the data as being an unknown classification based on the determination; and
perform an action set based on the unknown classification of the data.

2. The information handling system of claim 1 , wherein making the determination comprises:
obtaining a reconstruction error threshold for the decoder;
increasing the reconstruction error threshold based on the critical class classification indicating that the data is a member of a critical class detected by the critical class classifier to obtain a modified reconstruction error threshold;
making an identification that reconstruction error of the reconstructed data is larger than the modification reconstruction error threshold; and
making the determination based on the identification.

3. The information handling system of claim 1 , wherein making the determination comprises:
obtaining a reconstruction error threshold for the decoder;
making a first identification that the critical class classification indicates that the data is not a member of a critical class detected by the critical class classifier;
making a second identification that reconstruction error of the reconstructed data is larger than the reconstruction error threshold; and
making the determination based on the second identification.

4. The information handling system of claim 1 , wherein the processor is further programmed to:
obtain second data that may comprise the one or more of the objects;
encode the second data using the encoder to obtain second encoded data;
obtain a second critical class classification for the encoded second data using the critical class classifier;
obtain a second general classification for the encoded second data using the general classifier;
condition the encoded second data to obtain conditioned encoded second data;
decode the conditioned encoded second data to obtain second reconstructed data;
make a second determination that the second reconstructed data and the second critical class classification indicate that the second data is a known classification;
classify the data as being the second general classification based on the second determination; and
perform a second action set based on the second general classification of the second data.

5. The information handling system of claim 4 , wherein making the second determination comprises:
obtaining a reconstruction error threshold for the decoder;
increasing the reconstruction error threshold based on the second critical class classification indicating that the second data is a member of a critical class detected by the critical class classifier to obtain a modified reconstruction error threshold;
making an identification that reconstruction error of the second reconstructed data is larger than the modification reconstruction error threshold; and
making the second determination based on the identification.

6. The information handling system of claim 4 , wherein making the second determination comprises:
obtaining a reconstruction error threshold for the decoder;
making a first identification that the second critical class classification indicates that the second data is not a member of a critical class which the critical class classifier is trained to detect;
making a second identification that reconstruction error of the second reconstructed data is larger than the reconstruction error threshold; and
making the determination based on the second identification.

7. The information handling system of claim 1 , wherein the critical class classification for the encoded data is obtained by using the encoded data as input for the critical class classifier.

8. The information handling system of claim 1 , wherein the general classification for the encoded data is obtained by using the encoded data as input for the general classifier.

9. The information handling system of claim 1 , wherein conditioning the encoded data reduces a likelihood that reconstructed data will match the data when the data is dissimilar to training data used to obtain the decoder.

10. The information handling system of claim 1 , wherein the action set comprises:
providing the unknown classification of the data to an application that requested classification of the data.

11. A method for managing detection of objects, comprising:
obtaining data that may comprise one or more of the objects;
encoding the data using an encoder to obtain encoded data;
obtaining a critical class classification for the encoded data using a critical class classifier;
obtaining a general classification for the encoded data using a general classifier;
conditioning the encoded data to obtain conditioned encoded data;
decoding the conditioned encoded data using a decoder to obtain reconstructed data;
making a determination that the reconstructed data and the critical class classification indicate that the data is an unknown classification;
classifying the data as being an unknown classification based on the determination; and
performing an action set based on the unknown classification of the data.

12. The method of claim 11 , wherein making the determination comprises:
obtaining a reconstruction error threshold for the decoder;
increasing the reconstruction error threshold based on the critical class classification indicating that the data is a member of a critical class detected by the critical class classifier to obtain a modified reconstruction error threshold;
making an identification that reconstruction error of the reconstructed data is larger than the modification reconstruction error threshold; and
making the determination based on the identification.

13. The method of claim 11 , wherein making the determination comprises:
obtaining a reconstruction error threshold for the decoder;
making a first identification that the critical class classification indicates that the data is not a member of a critical class detected by the critical class classifier;
making a second identification that reconstruction error of the reconstructed data is larger than the reconstruction error threshold; and
making the determination based on the second identification.

14. The method of claim 11 , further comprising:
obtaining second data that may comprise the one or more of the objects;
encoding the second data using the encoder to obtain second encoded data;
obtaining a second critical class classification for the encoded second data using the critical class classifier;
obtaining a second general classification for the encoded second data using the general classifier;
conditioning the encoded second data to obtain conditioned encoded second data;
decoding the conditioned encoded second data using the decoder to obtain second reconstructed data;
making a second determination that the second reconstructed data and t

Assignees

• Emc Ip Holding Co Llc

Inventors

• Gottin Vinicius Michel
• Calmon Tiago Salviano
• Abelha Ferreira Paulo

Classifications

• G06F16/285 Primary

  Clustering or classification · CPC title

• G06F16/906 Primary

  Clustering; Classification · CPC title