aimagelab/meshed-memory-transformer
A transformer architecture with memory augmentation for generating textual descriptions from images, published at CVPR 2020.
Not currently ranked — collecting fresh signals.
star history
The Meshed-Memory Transformer (M2) is a deep learning model that generates captions for images using a modified transformer architecture. It introduces memory layers to enhance the model’s capacity for learning visual-semantic relationships. The model operates on pre-extracted detection features from a vision backbone and produces natural language descriptions through attention-based decoding. It was trained and evaluated on the COCO dataset.
Frequently asked
- What is aimagelab/meshed-memory-transformer?
- A transformer architecture with memory augmentation for generating textual descriptions from images, published at CVPR 2020.
- Is meshed-memory-transformer open source?
- Yes — aimagelab/meshed-memory-transformer is open source, released under the BSD-3-Clause license.
- What language is meshed-memory-transformer written in?
- aimagelab/meshed-memory-transformer is primarily written in Python.
- How popular is meshed-memory-transformer?
- aimagelab/meshed-memory-transformer has 546 stars on GitHub.
- Where can I find meshed-memory-transformer?
- aimagelab/meshed-memory-transformer is on GitHub at https://github.com/aimagelab/meshed-memory-transformer.