Flower Calvin Abcd
F
Flower Calvin Abcd
Developed by mbreuss
FlowerVLA is a robot operation model pre-trained on the CALVIN ABCD dataset, employing an innovative vision-language-action flow strategy with only 1 billion parameters, specifically designed for robot learning.
Downloads 24
Release Time : 3/16/2025
Model Overview
FlowerVLA is an efficient vision-language-action flow strategy that uses half-scale Florence-2 for multimodal vision-language encoding, combined with a novel Transformer-based flow matching architecture to achieve efficient and general-purpose vision-language-action policies.
Model Features
Efficient Multimodal Encoding
Uses half-scale Florence-2 for multimodal vision-language encoding to achieve efficient vision-language-action policies.
Innovative Flow Matching Architecture
Adopts a novel Transformer-based flow matching architecture with only about 1 billion parameters to achieve efficient and general-purpose vision-language-action policies.
High Performance
Ranked first in the CALVIN ABCD challenge with an average length of 4.72.
Model Capabilities
Vision-Language-Action Encoding
Robot Operation
Multimodal Task Execution
Use Cases
Robotics
Object Picking
Picks specific objects based on language instructions, such as a blue cube.
Achieved a 99.1% success rate in testing.
Featured Recommended AI Models