17879647. PICTURE QUALITY-SENSITIVE SEMANTIC SEGMENTATION FOR USE IN TRAINING IMAGE GENERATION ADVERSARIAL NETWORKS simplified abstract (Samsung Electronics Co., Ltd.)
PICTURE QUALITY-SENSITIVE SEMANTIC SEGMENTATION FOR USE IN TRAINING IMAGE GENERATION ADVERSARIAL NETWORKS
Organization Name
Inventor(s)
Hrishikesh Deepak Garud of Irvine CA (US)
PICTURE QUALITY-SENSITIVE SEMANTIC SEGMENTATION FOR USE IN TRAINING IMAGE GENERATION ADVERSARIAL NETWORKS - A simplified explanation of the abstract
This abstract first appeared for US patent application 17879647 titled 'PICTURE QUALITY-SENSITIVE SEMANTIC SEGMENTATION FOR USE IN TRAINING IMAGE GENERATION ADVERSARIAL NETWORKS
Simplified Explanation
The patent application describes a method for training a semantic segmentation network and utilizing it during the training of an image generation network. Here are the key points:
- The method involves training a semantic segmentation network to generate semantic segmentation maps with class-wise probability values.
- These semantic segmentation maps are then used during the training of an image generation network as part of a loss function that includes multiple losses.
- The semantic segmentation network is trained to be sensitive to the picture quality of the output image generated by the image generation network.
- If the picture quality of the output image degrades, the prediction confidence of the semantic segmentation network decreases.
- The semantic segmentation network can vary the class-wise probability values based on the picture quality.
Potential applications of this technology:
- Image generation and editing: The method can be used to improve the quality of generated or edited images by incorporating semantic segmentation maps into the training process.
- Computer vision tasks: The trained semantic segmentation network can be used for various computer vision tasks such as object detection, scene understanding, and image segmentation.
Problems solved by this technology:
- Picture quality degradation: By incorporating the semantic segmentation network's sensitivity to picture quality, the method helps in reducing the degradation of picture quality in generated images.
- Lack of context-awareness: The semantic segmentation maps provide contextual information to the image generation network, enabling it to generate more accurate and context-aware images.
Benefits of this technology:
- Improved image quality: By utilizing the semantic segmentation maps, the method can enhance the quality of generated images by considering both visual appearance and semantic context.
- Context-aware image generation: The image generation network can generate images that are more contextually relevant and aligned with the desired semantic segmentation.
- Efficient training: The method combines the training of the semantic segmentation network and the image generation network, leading to a more efficient and effective training process.
Original Abstract Submitted
A method includes training a semantic segmentation network to generate semantic segmentation maps having class-wise probability values. The method also includes generating a semantic segmentation map using the trained semantic segmentation network. The method further includes utilizing the semantic segmentation map during training of an image generation network as part of a loss function that includes multiple losses. The semantic segmentation network may be trained to be sensitive to picture quality of an output image generated by the image generation network during the training of the image generation network such that increased degradation of the picture quality of the output image results in decreased prediction confidence by the semantic segmentation network. The semantic segmentation network may be trained to vary the class-wise probability values based on the picture quality.
