PV-RNN visual predictions in response to language instructions

Visual representations of the different modules of the PV-RNN model as it predicts the visual output to a language instruction. The first clips show accurate predictions in response to “put green on blue”, whereas the latter show erroneous predictions to “put blue on yellow”.

Date:

19 December 2024

Creator:

adrian-skov

Credit:

Vijayaraghavan et al., 2025

Copyright OIST (Okinawa Institute of Science and Technology Graduate University, 沖縄科学技術大学院大学). Creative Commons Attribution 4.0 International License (CC BY 4.0).

Share on:

Related Videos

Gelation Video

This video shows a soapy mixture passing through the posts in the microfluidic platform and emerging on the right as a gel.

STEPS - Visualization toolkit

In the upper left box you can see all parts of the neuron in this model, the external structure and internal compartment (ER) that holds calcium (yellow), and the proteins involved in opening the calcium channel (blue, bright magenta and dark magenta). In the other boxes, you can see the calcium or each protein separately to watch their movements and activation. In the bottom center, you can see the numerical data in real time as the model runs.

X-Ray micro-CT scan of ant Pheidole fervens

X-Ray micro-CT scan of ant specimen Pheidole fervens by OIST Biodiversity & Biocomplexity Unit

Open Energy Systems Introduction (short)

This is the introduction movie of Open Energy Systems which are installed in 19 OIST faculty houses. Each house is connected with DC cable lines and exchange energy each other.

Higher Order Mode Video

This video shows a polystyrene particle propelled along a microfiber by light, first in the fundamental mode, then in a higher order mode. It is easy to see how much faster the particle zips along in the higher order mode.