A5067772383- Handwritten Text Recognition Techniques
- Vehicle License Plate Recognition
- Natural Language Processing Techniques
- Image Processing and 3D Reconstruction
- Machine Learning and Data Classification
- Anomaly Detection Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Electric Power System Optimization
- Medical Image Segmentation Techniques
- Smart Grid Security and Resilience
- Power System Reliability and Maintenance
- Medical Imaging Techniques and Applications
- Maritime Navigation and Safety
- Autonomous Vehicle Technology and Safety
- Data Stream Mining Techniques
- Smart Grid and Power Systems
- Energy Load and Power Forecasting
- Radiomics and Machine Learning in Medical Imaging
- Machine Learning and ELM
- Advanced Neural Network Applications
Tongji University
2024
Megvii (China)
2020
Shanghai Jiao Tong University
2018-2019
In this paper, we introduce a novel end-end framework for multi-oriented scene text detection from an instance-aware semantic segmentation perspective. We present Fused Text Segmentation Networks, which combine multi-level features during the feature extracting as instance may rely on finer expression compared to general objects. It detects and segments jointly simultaneously, leveraging merits both task region proposal based object task. Not involving any extra pipelines, our approach...
In this paper, we introduce a novel end-end framework for multi-oriented scene text detection from an instance-aware semantic segmentation perspective. We present Fused Text Segmentation Networks, which combine multi-level features during the feature extracting as instance may rely on finer expression compared to general objects. It detects and segments jointly simultaneously, leveraging merits both task region proposal based object task. Not involving any extra pipelines, our approach...
In this paper, we propose Double Supervised Network with Attention Mechanism (DSAN), a novel end-to-end trainable framework for scene text recognition. It incorporates one attention module during feature extraction which enforces the model to focus on regions and whole is supervised by two branches. One supervision comes from context-level modelling branch another extra enhancement aims at tackling inexplicit semantic information character level. These supervisions can benefit each other...
In this paper, we propose Double Supervised Network with Attention Mechanism (DSAN), a novel end-to-end trainable framework for scene text recognition. It incorporates one attention module during feature extraction which enforces the model to focus on regions and whole is supervised by two branches. One supervision branch comes from context-level modelling another extra enhancement aims at tackling inexplicit semantic information character level. These supervisions can benefit each other...
In this paper, we propose an end-to-end trainable framework for scene text spotting which can handle with arbitrary shapes. The proposed is called Word Segmentation Guided Characters Aggregation Net (WAC-Net), consists of a shared convolutional backbone and two task-specific subnetworks. One subnetwork does word-level instance-aware segmentation (WSN) the other char-level detection recognition (CDRN). entire segments each word instance while detects recognizes character in one single forward...
Knowledge distillation(KD) aims to improve the performance of a student network by mimicing knowledge from powerful teacher network. Existing methods focus on studying what should be transferred and treat all samples equally during training. This paper introduces adaptive sample weighting KD. We discover that previous effective hard mining are not appropriate for distillation. Furthermore, we propose Prime-Aware Adaptive Distillation (PAD) incorporation uncertainty learning. PAD perceives...