Brief Details: A specialized embedding model for Stable Diffusion v2 768 that creates fragmented artistic effects, particularly effective for character and object transformation.
Brief Details: Polish financial sentiment analysis model with 124M params, built on herbert-base. Achieves 0.969 F1-macro score for 3-class classification.
Brief-details: Vietnamese lyric alignment model based on wav2vec2, enables precise time-mapping of lyrics to audio using CTC-Segmentation. Achieves 0.632 IoU performance.
Brief-details: A specialized text-to-image model that generates Mr. Men and Little Miss style cartoon characters, built on Stable Diffusion with DreamBooth fine-tuning.
Brief-details: Experimental Dreambooth model generating 4x4 sprite sheets for abstract animations. Creates unique animated visuals at 512x512 resolution using specialized prompting.
Brief-details: T5-based Vietnamese text summarization model trained on news data. Optimized for 512 input/150 output tokens. MIT licensed, lightweight implementation.
Brief-details: Universal image segmentation model using DiNAT backbone, trained on COCO dataset. Performs semantic, instance & panoptic segmentation tasks with single architecture.
Brief-details: A Stable Diffusion fine-tuned model trained on personal artwork, offering character-focused image generation with 5k, 10k, and 15k step variants.
BRIEF DETAILS: Textual Inversion model for Stable Diffusion that generates Caitlin Fairchild character art in J. Scott Campbell's comic style, licensed under MIT.
Brief-details: T5-based model for converting UK English to US English, handling both spelling and vocabulary differences. Built on 264,519 sentence pairs.
Brief-details: Specialized Dreambooth model for generating Klonoa character images, trained on 814 pictures over 10k steps with hypernetwork enhancement. Stable Diffusion-based.
Brief-details: A Dreambooth model trained on personal artwork from Tumblr, featuring 1076 images trained over 10k steps. Combines anime and cartoon styles for text-to-image generation.
Brief Details: Turkish NLP transformer model with high-accuracy support for NER (91.3%), POS tagging (91.7%), and morphological analysis for the Turkish language
Brief Details: Multilingual text genre classifier supporting 94 languages, based on XLM-RoBERTa. 278M parameters, achieves 0.68 micro F1 score on cross-dataset evaluation.
Brief-details: miCSE - A 109M parameter sentence embedding model using mutual information contrastive learning, optimized for few-shot scenarios and achieving 78.13% avg on STS benchmarks.
Brief-details: Deep learning model for thyroid cancer analysis using H&E images. Predicts BRAF-RAS Score (BRS) for tumor classification. Built on Xception architecture with GPL-3.0 license.
Brief Details: Stable Diffusion v1.4 is a state-of-the-art text-to-image latent diffusion model, building upon v1.3 with enhanced aesthetics and classifier-free guidance capabilities.
Brief Details: MobileNetV2 with DeepLabV3+ for semantic segmentation, optimized for mobile devices. Pre-trained on PASCAL VOC at 513x513 resolution. Balances efficiency and accuracy.
Brief Details: Transform pop music into piano covers using T5-based transformer architecture. First model for direct audio-to-piano conversion without melody extraction.
Brief-details: BERT-based sarcasm detection model for English news headlines with 92.38% F1 score. Fine-tuned on Kaggle dataset, supports binary classification.
BRIEF DETAILS: Spanish legal domain Longformer model supporting 4,096 token sequences, based on RoBERTalex, trained on legal corpora with MIT license
