News

New fully open source vision encoder OpenVision arrives to improve on OpenAI’s Clip, Google’s SigLIP
A vision encoder is a necessary component for allowing many leading LLMs to be able to work with images uploaded by users.
Scientific Research Publishing3mon
A Combination Method of Stacked Convolutional Auto-Encoder and Selective Kernel Attention Mechanism for Image Classification - Scientific Research Publishing
Convolutional auto-encoders have shown their remarkable performance in stacking deep convolutional neural networks for classifying image data during the past several years. However, they are unable to ...
GitHub6mon
Hierarchical Encoder-decoder for Image Captioning - GitHub
The official repository for “Hierarchical Encoder-decoder for Image Captioning (HierCap)”. HierCap is a model to guide text generation with hierarchical visual information at three levels: global ...
IEEE8mon
A Comparative Evaluation of Transformer-Based Vision Encoder-Decoder Models for Brazilian Portuguese Image Captioning - IEEE Xplore
Image captioning refers to the process of creating a natural language description for one or more images. This task has several practical applications, from aiding in medical diagnoses through image ...
IEEE11mon
Image Caption Method from Coarse to Fine Based On Dual Encoder-Decoder Framework
In order to solve the above problems, we propose a coarse-fine image caption method based on dual encoder-decoder framework, which provides a mechanism for discovering and correcting omissions and ...
Frontiers1y
An image caption model based on attention mechanism and deep reinforcement learning
To address these issues, this paper studies and optimizes the image caption model with encoder-decoder architecture. The structure of the paper is arranged as follows: section 2 puts forward the image ...
marktechpost3y
Google AI Proposes Contrastive Captioner (CoCa): A Novel Encoder-Decoder Model That Simultaneously Produces Aligned Unimodal Image And Text Embeddings - MarkTechPost
On the other hand, encoder-decoder approaches are good at image captioning and visual question answering but not retrieval-style tasks. Google AI researchers provide a unified vision backbone model ...