Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/10—Segmentation; Edge detection
  • G06T7/11—Region-based segmentation
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/045—Combinations of networks
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/045—Combinations of networks
  • G06N3/0455—Auto-encoder networks; Encoder-decoder networks
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/0464—Convolutional networks [CNN, ConvNet]
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods
  • G06N3/084—Backpropagation, e.g. using gradient descent
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods
  • G06N3/09—Supervised learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods
  • G06N3/096—Transfer learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10016—Video; Image sequence
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20081—Training; Learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20084—Artificial neural networks [ANN]
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30248—Vehicle exterior or interior
  • G06T2207/30252—Vehicle exterior; Vicinity of vehicle
  • G06T2207/30261—Obstacle

Definitions

• the present disclosure relates to image processing and, more particularly, to methods of image processing using artificial intelligence.
• the present disclosure relates to a system that includes at least one memory device and at least one processing device coupled at least indirectly with the at least one memory device.
• the at least one processing device is (or are) configured to operate in accordance with a convolutional neural network architecture including an encoder network, a semantic segmentation decoder network, and an instance segmentation decoder network.
• the encoder network and the semantic segmentation decoder network are configured to train on a dataset to perform a semantic segmentation task.
• the encoder network is further configured to generate first feature maps during the semantic segmentation task.
• the instance segmentation decoder network is configured to receive the first feature maps generated by the encoder network.
• the instance segmentation decoder network is configured to train using the first feature maps and the dataset to perform an instance segmentation task.
• FIG. 1 is a schematic diagram showing an example electrical system implemented in part on a vehicle in accordance with an example embodiment encompassed herein;
• the instance segmentation decoder network 306 can comprise a detector subnet and a segmentation subnet.
• the detector subnet can comprise a bounding box regressor and a classifier, as would be understood by those having skill in the art.
• the instance segmentation decoder network 306 can comprise Mask R-CNN, which was introduced by He et al (Kaiming He, Georgia Gkioxari, Piotr Dollar, and Ross B. Girshick. Mask R-CNN. CoRR, abs/1703.06870, 2017, which is incorporated by reference herein in its entirety).
• instance segmentation comprises a more difficult task involving higher reasoning of machine learning than semantic segmentation
• the more difficult task can learn from representations (in a form of feature maps) learned from the easier task (semantic segmentation).
• parameters from the encoder, semantic segmentation decoder network, and the instance segmentation decoder network can be combined to initialize the multi-task convolutional neural network.
• the encoder network does not train with the instance segmentation decoder network, but instead provides outputs (e.g., feature maps) generated when performing semantic segmentation with the semantic segmentation decoder network.
• the output provided by the encoder network 302 to the instance segmentation decoder network 306 can take the form of feature maps, which can be of the same type as the feature maps 614 generated by the encoder fine tuned in the first stage 610 and provided for use in the second stage 620.
• any feature maps generated by the encoder network 302 in the third stage 630 are generated in response to the image 660, which is different from the dataset (image) 640 received by the encoder network 302 in the first stage 610.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• General Physics & Mathematics (AREA)
• General Health & Medical Sciences (AREA)
• Molecular Biology (AREA)
• Biophysics (AREA)
• Computational Linguistics (AREA)
• Data Mining & Analysis (AREA)
• Evolutionary Computation (AREA)
• Artificial Intelligence (AREA)
• Biomedical Technology (AREA)
• Computing Systems (AREA)
• General Engineering & Computer Science (AREA)
• Life Sciences & Earth Sciences (AREA)
• Mathematical Physics (AREA)
• Software Systems (AREA)
• Health & Medical Sciences (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Image Analysis (AREA)

Applications Claiming Priority (1)

Publications (2)

Family

ID=71407381

Family Applications (1)

Country Status (2)

Cited By (2)

Families Citing this family (15)

Family Cites Families (1)

• 2018
  • 2018-12-31 EP EP18913207.9A patent/EP3698286A4/de not_active Withdrawn
  • 2018-12-31 WO PCT/US2018/068172 patent/WO2020142077A1/en not_active Ceased

Also Published As

Similar Documents

Legal Events