Pinned Repositories
Abstractive-Summarization-RoBERTa2RoBERTa
Leveraging RoBERTa for Abstractive Summarization using Pre-trained Encoders
AnomalyDetection
In this study, we introduce a robust baseline for real-time, transformer-based, object-centric video anomaly detection using YOLOv8 and RAFT. Our approach, tested on Avenue, ShanghaiTech, and UBNormal datasets, is competitive with frame-level methods in accuracy and more efficient than other object-centric methods.
Binarized-Neural-Machine-Translation
Results reproduction of the paper named Binarized Neural Machine Translation, where we leverage quantized transformer architecture for a more computationally efficient model.
DL-Synthetic-Image-Classification-Competition
Placed 2nd (out of 65) in the Deep Learning course Kaggle Competition, where all students in the second year of University of Bucharest AI MSc would compete against each other for best model score. Dataset was composed of synthetic/generated images and the purpose was to classify images into 100 classes.
GroundingDINO
Official implementation of the paper "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection"
Leveraging-BERT-for-Emotion-Cause-Pair-Detection-in-Conversational-Contexts
Leveraging BERT for Enhanced Sentiment Analysis and Emotion-Cause Pair Detection in Conversational Contexts
mae
PyTorch implementation of MAE https//arxiv.org/abs/2111.06377
ml-fastvit
This repository contains the official implementation of the research paper, "FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization" ICCV 2023
OW-DETR
[CVPR 2022] Official Pytorch code for OW-DETR: Open-world Detection Transformer
TruthDetection
My BSc Thesis implementation which dwells upon the complex task of Truth Detection (often also referred to as Deception Detection or Lie Detection).
the-sergiu's Repositories
the-sergiu/TruthDetection
My BSc Thesis implementation which dwells upon the complex task of Truth Detection (often also referred to as Deception Detection or Lie Detection).
the-sergiu/AnomalyDetection
In this study, we introduce a robust baseline for real-time, transformer-based, object-centric video anomaly detection using YOLOv8 and RAFT. Our approach, tested on Avenue, ShanghaiTech, and UBNormal datasets, is competitive with frame-level methods in accuracy and more efficient than other object-centric methods.
the-sergiu/GroundingDINO
Official implementation of the paper "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection"
the-sergiu/mae
PyTorch implementation of MAE https//arxiv.org/abs/2111.06377
the-sergiu/ml-fastvit
This repository contains the official implementation of the research paper, "FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization" ICCV 2023
the-sergiu/OW-DETR
[CVPR 2022] Official Pytorch code for OW-DETR: Open-world Detection Transformer
the-sergiu/recognize-anything
Code for the Recognize Anything Model (RAM) and Tag2Text Model
the-sergiu/Visual-Traffic-Monitoring-Computer-Vision-Challenge
The goal of this project is to develop an automatic system for video analysis that enables visual traffic monitoring at a road intersection . The system gets as input data the video stream from a static camera and should be able to track cars, classify lanes, and determine trajectory of vehicles.
the-sergiu/Abstractive-Summarization-RoBERTa2RoBERTa
Leveraging RoBERTa for Abstractive Summarization using Pre-trained Encoders
the-sergiu/Binarized-Neural-Machine-Translation
Results reproduction of the paper named Binarized Neural Machine Translation, where we leverage quantized transformer architecture for a more computationally efficient model.
the-sergiu/DL-Synthetic-Image-Classification-Competition
Placed 2nd (out of 65) in the Deep Learning course Kaggle Competition, where all students in the second year of University of Bucharest AI MSc would compete against each other for best model score. Dataset was composed of synthetic/generated images and the purpose was to classify images into 100 classes.
the-sergiu/Fine-Tuning-Stable-Diffusion-for-Cartoon-Generation
Fine-tuned Stable Diffusion v1.5 (InvokeAI codebase) using processed images in the style of South Park, Simpsons and Naruto. Created realistic images, abstract art and combined these styles which resulted in new types of cartoons
the-sergiu/Leveraging-BERT-for-Emotion-Cause-Pair-Detection-in-Conversational-Contexts
Leveraging BERT for Enhanced Sentiment Analysis and Emotion-Cause Pair Detection in Conversational Contexts
the-sergiu/Arithmetic-Logical-Unit-in-Verilog
Task was to implement the functionalities of an ALU, based on a master-slave principle. The code was based upon the Intel Avalon Memory scheme (slave), with the ALU being the master. The ALU had to fulfill basic functionalities/operations, such as ADD, AND, OR, XOR, DEC, INC, SHL, SHR, with different types of memory addressing and varying number of operands. Implemented in Verilog, using Xilinx IDE.
the-sergiu/C-Language-Preprocessor
This is a simple C/C++ preprocessor. The goal is to have good conformance with the C and C++ standards and to handle nonstandard preprocessor extensions in gcc / clang / visual studio preprocessors. Most of the preprocessor testcases in gcc and clang are handled decently by so-cpp.
the-sergiu/Checker_eStorePlatform
Automated-checking script for the eStore Platform Assignment. Used as Assignment for Object-Oriented Programming (C++) at the Faculty of Automatic Control and Computer Science @ Politehnica University of Bucharest, 2020-2021. Valued at 15% of the entire grade.
the-sergiu/conversational-agent
the-sergiu/CVND_Exercises
Exercise notebooks for CVND.
the-sergiu/detect-components-document
the-sergiu/DNA-Diffusion
💨 Understanding the code of life: Generative modeling of regulatory DNA sequences with diffusion probabilistic models 🧬
the-sergiu/Double-Double-Domino-Computer-Vision-Challenge
The goal of this project is to develop an automatic system for scoring a specific variant of dominoes game called Double Double Dominoes.
the-sergiu/eStore_Platform
As the name suggests, we wish to create and develop the back-end of an e-store/e-commerce app with the purpose of selling various products. Used as Assignment for Object-Oriented Programming (C++) at the Faculty of Automatic Control and Computer Science @ Politehnica University of Bucharest, 2020-2021.
the-sergiu/Executable-File-Bootloader-in-C
The task was to implement an Executable File Bootloader as a dynamic library for UNIX. The loader would load the binary executable in the main memory, page by page, using the demand-paging mechanism.
the-sergiu/Google-Speech-To-Text-API-Streamlit-Python
Streamlit App that lets you intuitively convert waveform files to text, in English and Romanian. Google Cloud API Credentials Request JSON required via Google Cloud Credentials.
the-sergiu/Housing_Prices_Advanced_Regression_Kaggle_Competition
Predict sales prices and practice feature engineering, RFs, and gradient boosting. Top 13% entries as of June 2021.
the-sergiu/minitorch
A smaller implementation of PyTorch which contains most of the essential components.
the-sergiu/PML-Smarthphone-User-Identification-Competition
Placed 3rd (out of 80) in the Practical Machine Learning course Kaggle Competition, where all students in first year of AI MSc would compete against each other for best model score. Dataset was composed of accelerometer timeseries on multiple axis and the purpose was to classify users based on behavior.
the-sergiu/Standard-Input-Output-Library-in-C
The so_stdio.h header defines three variable types, several macros, and various functions for performing buffered input and output operations with files. Implemented in C for POSIX and WIN32 API.
the-sergiu/the-sergiu
the-sergiu/Waveform-Spectrogram-Vizualizer-in-Streamlit
This is a Streamlit App that helps compare and visualize two Waveform (WAV) files from a signal processing point of view.