IEEE

TrOCR: Transformer-based OCR with Pre-trained Vision-Language Models

Abstract: Transformer-based models have revolutionized Optical Character Recognition (OCR) by enabling end-to-end learning from raw image inputs to textual outputs. TrOCR, a recent innovation, ...
IEEE

Secure Transfer Learning: Training Clean Model Against Backdoor in Pre-Trained Encoder and Downstream Dataset

Abstract: Transfer learning from pre-trained encoders has become essential in modern machine learning, enabling efficient model adaptation across diverse tasks. However, this combination of ...
GitHub

Vision Foundation Models as Effective Visual Tokenizers for Autoregressive Generation

This is a PyTorch/GPU implementation of the paper Vision Foundation Models as Effective Visual Tokenizers for Autoregressive Generation, which directly utilizes the features from the frozen ...
GitHub

building_robust_vision_encoders_for_cross-dataset_evaluation_in_immunofluorescen.md

Architecturally, groups are encoded independently and then fused (CCE). During pre-training, the concept channels learn to contribute to representations even under "limited context" (MCD). At ...