Joy Caption is a ComfyUI node using the LLaVA model to generate stylized image captions, supporting batch processing and GGUF models.

Source code for paper "Multi-Classification In-Vehicle Intrusion Detection System using Packet- and Sequence-Level Characteristics from Time-Embedded Transformer with Autoencoder"

Cross-platform FlashAttention-2 Triton implementation for Turing+ GPUs with custom configuration mode

This repository contains code for generating captions for images using a Transformer-based model. The model used is the `VisionEncoderDecoderModel` from the Hugging Face Transformers library, specifically the `nlpconnect/vit-gpt2-image-captioning` model.

A Next-Word Prediction project uses Transformers and GPT-2 for text generation. GPTTokenizer preprocesses input, and the model is fine-tuned. Evaluation measures accuracy, perplexity, and fluency.

Visual-Grounding-Anything is a comprehensive suite of applications designed for precise object detection, pointing, and tracking in both images and videos. Leveraging the Polaris-VGA-4B model, the system provides high-accuracy spatial reasoning and temporal association across various visual tasks.

Incredibly fast Whisper-large-v3 with speaker diarization

A constitutional system for human-AI collaboration, fostering a practical, ethical, scalable, and sustainable framework with AI as a critical peer. Governed by principles of shared responsibility, guardrails, and rapid evolution, informed by human-centered AI, IaC, CaC, and transformers.

Speech-to-Text (STT) with Whisper

A simple book recommender project with a Python backend and a Next.js frontend.

Sentiment analysis project using VADER and RoBERTa to classify text into positive, negative, and neutral sentiments.