Codes and Datasets

Fuzzy-Discernibility-Matrix-v2.0.0

The new Fuzzy Discernibility Matrix-based can also work with continuous value dataset: my_fuzzy_discrn_mat.py (for any number of features). Example: Irish_Evaluation_notebook.ipynb demonstrates both the feature selection techniques on standard Iris dataset.

🔗 GitHub Repo

Dataset: Iris or any numerical dataset

YOLOv13-Fine-Tune-on-Custom-Data

YOLOv13 Fine-tune Object Detection.

🔗 GitHub Repo

Dataset: Roboflow: BCCD-4

LIQUID NEURAL NETWORK LNN

The code is now much more robust, handles edge cases better, and should provide significantly better performance on both classification and regression tasks.

🔗 GitHub Repo

Datasets Used:

PyCaret-and-Titanic-Dataset

The code is to provide a hands-on experience about EDA on any dataset using python library PyCaret.

🔗 GitHub Repo

Dataset: Titanic dataset

VIA-COCO Converter

This repo. is meant for those who find difficulty in dealing with formatting issues with annotations (Faster RCNN, Mask RCNN, etc.) e.g., VGG-Image-Annotator (VIA) JSON to COCO JSON, JSON to XML, and XML to TXT, vice versa.

🔗 GitHub Repo

No dataset specified

NASA-Image-Downloader

One can get a NASA API key at https://api.nasa.gov/. Then, save your key as an environment variable called NASA_API_KEY.

🔗 GitHub Repo

No dataset specified

ELM-Python

The code is python adaptation of Extreme Learning Machine.

🔗 GitHub Repo

Dataset: Iris dataset (work with any classification task)

Drones in Defence: Military Target Surveillance and Tracking

This study presents a drone-captured video dataset for military surveillance and evaluates YOLOv5, YOLOv8, and YOLOv11 models with their variants to balance accuracy and efficiency. The results show that YOLOv11s combined with ByteTrack achieves the best performance for real-time detection and tracking of enemy assets, advancing autonomous military surveillance.

🔗 GitHub Repo

Dataset: KIIT-MiTA

ClipXpert: Automated Clip Mining from Video Data for High-Demand Content

ClipXpert is an automated system that extracts relevant YouTube clips using keywords, frequent word/comment analysis, and transcription models, while dynamically storing new content for future use. It further identifies key linguistic features and applies sentiment analysis on comments to generate focused highlights, improving efficiency and accuracy in targeted video content extraction.

🔗 GitHub Repo

Dataset:

South Asian Sounds: Audio Classification

Sound classification plays a vital role in urban planning and noise monitoring. Leveraging MFCCs and a 1D-CNN, we classify sounds from Bangladesh, India, and the UrbanSounds8k dataset. Our model demonstrates high accuracy, highlighting the effectiveness of combining MFCCs with 1D-CNNs for urban sound classification.

🔗 GitHub Repo

Dataset: SAS-KIIT