Abstract: Unsupervised anomaly detection (UAD) aims to recognize anomalous images based on the training set that contains only normal images. In medical image analysis, UAD benefits from leveraging ...

We propose an encoder-decoder for open-vocabulary semantic segmentation comprising a hierarchical encoder-based cost map generation and a gradual fusion decoder. We introduce a category early ...

IDProtector is a model that adds protective noise to images, such that if the protected image is used in encoder-based identity-preserving generation, the resulting face will no longer resemble the ...

Abstract: To improve the measurement precision of the optical encoder working at various speeds, an adaptive high-precision measurement framework is designed based on deep reinforcement learning (DRL) ...