Adapting VLMs such as CLIP to specific Low-Resources tasks in a Few Shot Learning (FSL) context.
Few Shot Learning
Adapters
PyTorch
TTA : Test Time Adaptation
Implementation of a TTA solution to increase image classification accuracy without re-training a model
Deep Learning
TTA
CLIP
PyTorch
Robotic manipulator
From the object detection and pose estimation of rigid objects to evaluating the dynamics and trajectories of a 6-DOF UR5 manipulator in order to achieve automatically "pick and place" tasks
ROS
Object detection
Pose estimation
Artificial Dataset generation
Motion capturing
An experiment about manipulating and correcting motion capture data from an optitrack system
MoCap
Unreal Engine 5
3d environments
CUDA programming
A dive into GPU programming with CUDA, tackling problems such as the matrix transposition and image processing with the convolution operation
GPU programming
Parallel computation
CUDA
C
Other topics I've worked on
Volleyball ball-tracking
A computer vision & vanilla machine learning application which aims to track a ball in a Volleyball game scenario.
