A comprehensive Streamlit web application for analyzing cryptocurrency prices and predicting next-day HIGH prices using advanced machine learning models. View real-time historical data with interactive candlestick charts and get AI-powered price predictions for Bitcoin, Ethereum, XRP, and Solana.

github: https://www.twine.net/signin
Api: https://www.twine.net/signin
Live Demo: https://www.twine.net/signin

Features

Real-Time Price Charts: Interactive candlestick charts with 30-day historical OHLC (Open, High, Low, Close) data from Kraken API
AI Price Predictions: Machine learning-powered predictions for next-day HIGH prices
Multi-Cryptocurrency Support: Bitcoin (BTC), Ethereum (ETH), XRP, and Solana (SOL)
Price Metrics: Current price, 24-hour change, 30-day high, and predicted returns
Professional UI: Clean, intuitive interface with cryptocurrency logos and icons
Fast & Responsive: Built with Streamlit for smooth, real-time updates
Automatic Data Fetching: All prediction APIs automatically fetch latest market data
Multiple Display Formats: Each cryptocurrency shows predictions in its optimal format

Accurate Vehicle Identification plays a crucial role in today’s law enforcement and public safety. NSW Police purely depend on surveillance video footage for the investigation of crime incidents, tracking of suspected vehicles to gather evidence(Dowling et al., n.d.). But the current systems that are dependent mainly on number plate recognition fail when plates are obscured, damaged or replaced(Lubna et al., n.d.). This highlights the need for advanced technology in identifying the vehicles based on the vehicle physical characteristics rather than number plate registration(Karrach et al., 2020; Risha et al., 2025).
Our Project, Vehicle Metadata Identification Using Machine Learning, will address these issues by developing an automated image-based system which will be capable of recognising key attributes of a vehicle such as make, model, body type, colour and visible accessories or damage. The motivation behind the research is to enhance investigative efficiency and accuracy by reducing the manual effort required to review video footage(Nilsson, 2023; Semiotics & 2013, 2013).
To accomplish this goal, a multimodal machine learning pipeline was modelled and used. The method has multiple stages, including vehicle discovery utilizing the YOLOv8 structure to localize cars in pictures, classification models to recognize adequate metadata characteristics(Arjun et al., 2025; Sapkota, Flores-Calero, Qureshi, Badgujar, et al., 2025). In addition, CNNs, EffectiveNet, and Vision Transformers were examined as common spines for functionality extraction(Nejad et al., 2023). Additional specific model heads distinguishing responsibilities used where diverse neuronal subnets were required(Yuan et al., 2022). Models were taught on publicly accessible data sets such as CompCars and Stanford Cars(Gayen et al., 2025; Munoz et al., 2025).
Performance was evaluated using measures such as accuracy, precision, recall, F1-score, and mean Average Precision. As proved by the results, the modular approach proposed helped increase the dependability of vehicle identification and its readiness for scaling even when small fragments of cars are visible due to lighting conditions. Moreover, the modular approach improved the interpretability of the model proposed, as it can be used to reshape and rebuild the components separately when newer technologies are invented(Sharma et al., 2021).
This project does a lot. We’ve got cleaned and prepped datasets, trained models, performance reports, and even a working demo tool. All of it shows how machine learning can really help law enforcement - faster, smarter, and automated vehicle recognition. The same approach can be used for traffic management, insurance checks, or smarter transportation systems. In the end, the project pushes computer vision forward and proves that AI isn’t just hype - it actually makes a difference for public safety and city life.