Brief Details: Efficient sentence embedding model that maps text to 384-dimensional vectors. Based on MiniLM architecture, trained on 1B+ sentence pairs for semantic search and clustering.
Brief Details: A Kabyle language ASR model using wav2vec 2.0 with CTC/Attention, achieving 24.80% WER on test data. Built by AIOX Labs using SpeechBrain.
Brief-details: A fine-tuned wav2vec2-XLS-R speech recognition model with 315M parameters, achieving 13.32% CER, optimized for phoneme recognition using PyTorch.
Brief Details: Korean T5 language model with 820M parameters, trained on Korean text using BBPE tokenization. Optimized for text generation and transfer learning tasks.
Brief-details: PIXEL (Pixel-based Encoder of Language) is an innovative 86M parameter language model that processes text as rendered images rather than tokens, trained on 3.2B words.
Brief Details: BiodivBERT is a specialized BERT-based model for biodiversity literature, supporting NER and relation extraction tasks with state-of-the-art performance.
Brief-details: Financial sentiment analysis model fine-tuned from FinancialBERT achieving 99.24% accuracy on financial_phrasebank dataset. Optimized for text classification.
BRIEF DETAILS: Multi-task RoBERTa-based model for named entity recognition, relation extraction, and coreference resolution. Supports 18 entity types and 5 relation types.
BRIEF DETAILS: FinBERT-ESG: BERT-based model for ESG text classification in financial documents. Fine-tuned on 2,000 annotated sentences for environmental, social, and governance analysis.
Brief-details: A specialized BERT model for biomedical text and chest X-ray analysis, pretrained on PubMed and MIMIC datasets, achieving SOTA performance in radiology NLP tasks.
Brief Details: T5-3B-based model for converting question-answer pairs into declarative statements. Enables natural language processing tasks with 40+ downloads.
Brief Details: Vietnamese speech recognition model pre-trained on 3K hours of audio data. 95M parameters, supports 16kHz audio input, built on wav2vec2 architecture.
Brief-details: Multilingual intent classification model supporting 51 languages, trained on MASSIVE dataset for Alexa-like commands with 60 different intents.
Brief-details: ESPnet-based Persian ASR model trained on CommonVoice dataset, achieving 91.4% WER and 97.2% CER using BLSTM architecture
Brief-details: A BERT-based sentiment analysis model fine-tuned on Amazon reviews, achieving 80% accuracy for 1-5 star rating predictions across multiple languages.
Brief-details: LayoutLMv3 model fine-tuned on FUNSD dataset for document understanding, achieving 90.78% F1 score. Specialized in token classification tasks with visual-language capabilities.
Brief Details: Spanish doc2query model based on mT5, specialized in document expansion and query generation for improved search relevance and training data creation.
Brief-details: Multilingual text summarization model that converts text from 43 languages into Chinese (Simplified) summaries using mT5 architecture
BRIEF-DETAILS: NeonGAN is a CycleGAN-based model that transforms regular images into futuristic neon-style versions, built with PyTorch and trained on 256x256 images.
Brief-details: ResNet-50 is a powerful image recognition model implementing deep residual learning architecture, optimized for computer vision tasks with ONNX runtime support.
Brief-details: RoBERTa-based fake news detection model trained on 40k+ news articles. Classifies text as real/fake with high accuracy. Apache 2.0 licensed.
