projects

research
project thumbnail

Wrapyfi

Wrapyfi is a middleware communication wrapper for transmitting data across nodes, without altering the operation pipeline of your Python scripts. Wrapyfi introduces a number of helper functions to make middleware integration possible without the need to learn an entire framework, just to parallelize your processes on multiple machines. Wrapyfi supports YARP, ROS, ROS 2, and ZeroMQ.
project thumbnail

GASP

GASP is a multimodal social cue integration model for dynamic (video) saliency prediction. It combines the representations of existing social cue detectors (gaze estimation, gaze following, facial expression recognition, and audiovisual saliency prediction) and attends to the most conspicious region.
project thumbnail

Prompt-Driven Visuomotor Grasping

A neural model for grasping basic objects with the NICO robot. The initial and target images are fed into the model, along with the shape, color and position of the object to be grasped, and where it should be placed.

OSS
project thumbnail

ImageBind LoRA Fine-Tuning

Fine-tuning "ImageBind One Embedding Space to Bind Them All" using Low Rank Adaptation (LoRA)
project thumbnail

Distributing the Llama LLM with Wrapyfi

Wrapyfi enables distributing LLaMA (inference only) on multiple GPUs/machines, each with less than 16GB VRAM

datasets
project thumbnail

Augmented Extended Train Robots Dataset

A dataset containing 3D augmented images of basic shapes organized based on the ETR dataset. The dataset contains instructions for moving a block from one position to another. Basic linguistic instructions, RCL commands, and speech (WaveNet TTS) are provided.

contributions

6DRepNet

Refactored repository for packaging source and correctly linking imports.

al-folio

Added support for video blocks in bib entries.

goal_conditioned_RL_baselines

Added multiple robotic platforms (physical- pianobot. simulated- NICO robot), created a modular generalized model interface, and modified simulation environments.