Location: Warsaw University of Technology, Poland Website: cvlab.ii.pw.edu.pl Available projects: 9
Computer Vision Lab is a research group at Warsaw University of Technology that gathers faculty and students working together on topics at the intersection of computer vision, machine learning and perception.
Point clouds from Time-Of-Flight (ToF) cameras, rendered from depth images in the infrared (IR) band, are susceptible to reflective elements in the environment. The project aims to develop a system identifying anomalies in depth images or point clouds, enabling data filtering from the sensor.
Contact: Tomasz Trzcinski (konrad.cop.dokt [ at ] pw.edu.pl)
See project's pageTransformers are the foundation for many well performing neural language processing models. Unfortunately, they require a lot of computational resources which results in slow inference. In this project we aim to leverage conditional computation methods to speed up inference along three axes: depth-wise sparsity (early exits), width-wise sparsity (mixture of experts) and input-wise sparsity (dynamic sequence pruning). Additionally, we would like examine the hypothesis that some data points are easier to process for neural networks. For that purpose, among others, we would like to implement dynamic variant of mixture of experts (MoE) that enables MoE layers to use less resources for easy data points and compare it with difficulty rating extracted from early exit models.
Contact: Tomasz Trzcinski (filip.szatkowski [ at ] pw.edu.pl)
See project's pageThe project will focus on developing a method to reduce forgetting in neural networks using feature replay techniques.
Contact: Tomasz Trzcinski (stanislaw.pawlak.dokt [ at ] pw.edu.pl)
See project's pageThe project focuses on analyzing knowledge transfer and forgetting in hierarchical decision systems in continual learning scenarios.
Contact: Tomasz Trzcinski (michalbortkiewicz8 [ at ] gmail.com)
See project's pageThe project aims to develop a system for detecting grates in point clouds, RGB, or depth images and classifying them as safe for robot navigation.
Contact: Tomasz Trzcinski (konrad.cop.dokt [ at ] pw.edu.pl)
See project's pageThe goal of the project is to develop a method to mitigate the phenomenon of forgetting by utilizing various noise generation techniques. The starting point for further research will be the paper Learning to See by Looking at Noise. The work will encompass a review of currently applied rehearsal methods and the creation and in-depth analysis of a method utilizing noise to reduce forgetting. Two key aspects of the work will be optimizing the method for the model’s ultimate resistance to forgetting and the computational complexity of the proposed method.
Contact: Tomasz Trzcinski (wojciech.masarczyk.dokt [ at ] pw.edu.pl)
See project's pageThe goal of the project is to develop an automatic method for constructing quantum circuits. To execute a given program on a quantum computer, it must be expressed using a quantum circuit, i.e., a sequence of quantum gates. The project proposes a method based on the AlphaGo algorithm to develop a strategy for constructing quantum circuits.
Contact: Tomasz Trzcinski (mm.ostaszewski [ at ] gmail.com)
See project's pageThe objective of the project is to develop a system for performing ultrasonographic examinations based on reinforcement learning and a collaborative robot arm xArm7. The solution involves the use of deep convolutional neural networks and computer vision algorithms to train the collaborative robot for automatic ultrasound examinations using an ultrasound probe. The project is being carried out in collaboration with the Sano computational medicine research institute. There is an opportunity to receive a scholarship.
Contact: Tomasz Trzcinski (s.plotka [ at ] sanoscience.org)
See project's pageThe project aims to implement a bacterial growth classification algorithm in Scope Fluidics’ BacterOMIC machine with a focus on visual transformers and data-centric AI.
Contact: Tomasz Trzcinski (michalbortkiewicz8 [ at ] gmail.com)
See project's page