Avatar

Agnieszka Mikołajczyk

PhD Candidate & Machine Learning Enthusiast

Gdańsk University of Technology

Biography

PhD Agnieszka Mikołajczyk is an author of datasets, scientific papers, and publications, holding numerous scholarships and awards. She works on Large Language Models like the first Polish GPT model called TRURL. She is a co-organizer of PolEval2021 and PolEval 2022 tasks with punctuation prediction and restoration. She organizes and actively contributes to the scientific community in her free time: she managed and led the team during the HearAI project focused on modeling Sign Language (hearai.pl). A former organizer and a team leader at the open-source project detectwaste.ml. As an ML Expert, she supports the project “Susana” designed to detect and read product expiry dates to help the Blind “see”. Daily, she conducts her research on her grant “Detecting and overcoming bias in data with explainable artificial intelligence” Preludium, awarded by Polish National Centre (biasinml.netlify.app).

Interests

  • Explainable Artificial Intelligence
  • Image Analysis
  • Deep Learning

Education

  • PhD in Machine Learning, 2017 - 2021

    Gdańsk University of Technology

  • MEng in Control Theory, 2016 - 2017

    Gdańsk University of Technology

  • BSc in Automation Control and Robotics, 2012 - 2016

    Gdańsk University of Technology

Projects

.js-id-projects

Detecting and Reducing Bias in Data

Currently, in contrast to shallow models exploited in the past, most deep learning systems extract features automatically, and to do …

DetectWaste and ClassifyWaste

The proposed classify-waste benchmark is a merged collection of publicly available datasets with eight classification labels. The …

HearAI

Deaf people are affected by many forms of exclusion, especially now in the pandemic world. HearAI aims to build a deep learning …

Punctuation Restoration

Speech transcripts generated by Automatic Speech Recognition (ASR) systems typically do not contain any punctuation or capitalization. …

Tiny Hero - Generating pixel characters with GANs

Dataset TinyHero includes 64x64 retro-pixel character. All characters were generated with Universal LPC spritesheet by makrohn. Each …

Skin Lesion Classification

In the last twenty years the interest of automated skin lesion classification dynamically increased partially because of public …

Bird Song Classification

Sound-Based Bird Classification using Convolutional Neural Networks and Mel-Cepstrum Sepctrograms

Detect waste in Pomerania

Using detection models to localize and classify waste on images and video.

Hack4Environment

Hackathon. Let's do something for our environment.

Datasets

ML data collections

.js-id-datasets

Detecting and Reducing Bias in Data

Currently, in contrast to shallow models exploited in the past, most deep learning systems extract features automatically, and to do …

DetectWaste and ClassifyWaste

The proposed classify-waste benchmark is a merged collection of publicly available datasets with eight classification labels. The …

HearAI

Deaf people are affected by many forms of exclusion, especially now in the pandemic world. HearAI aims to build a deep learning …

Punctuation Restoration

Speech transcripts generated by Automatic Speech Recognition (ASR) systems typically do not contain any punctuation or capitalization. …

Tiny Hero - Generating pixel characters with GANs

Dataset TinyHero includes 64x64 retro-pixel character. All characters were generated with Universal LPC spritesheet by makrohn. Each …

Skin Lesion Classification

In the last twenty years the interest of automated skin lesion classification dynamically increased partially because of public …

Bird Song Classification

Sound-Based Bird Classification using Convolutional Neural Networks and Mel-Cepstrum Sepctrograms

Detect waste in Pomerania

Using detection models to localize and classify waste on images and video.

Hack4Environment

Hackathon. Let's do something for our environment.

Recent & Upcoming Talks

AI, Acoustics & Ornithology for sound-based bird classification

Have you ever wondered about the name of the bird you just heard singing? A group of women from local Polish chapter of Women in …

Introduction to Explainable AI: why should we understand AI decisions?

Is XAI really that important? Why should we try to explain our models’ predictions? A short introduction to explainable AI.

Brief introduction to XAI

Brief introduction to visual local explainabiliy methods - Local Interpretable Model-Agnostic Explanations (LIME), Layer-wise Relevance …

Recent Publications

(2022). Deep learning-based waste detection in natural and urban environments. Waste Management.

PDF

(2021). A Comprehensive Analysis of Deep Neural-Based Cerebral Microbleeds Detection System. Electronics.

PDF

(2021). Joint prediction of truecasing and punctuation for conversational speech in low-resource scenarios. arXiv preprint arXiv:2109.06103.

PDF

(2021). Towards explainable classifiers using the counterfactual approach-global explanations for discovering bias in data. Journal of Artificial Intelligence and Soft Computing Research.

PDF

(2021). Waste detection in Pomerania: non-profit project for detecting waste in environment. arXiv preprint arXiv:2105.06808.

PDF

Contact

  • Gabriela Narutowicza 11/12, Gdańsk, GDA 80244
  • Enter faculty of Electrical and Control Engineering Building and take the stairs to Office 207 on Floor 2