# Zero-shot transfer

Vision-language model fine-tuned based on CLIP-ViT-B/32, suitable for image-text matching tasks

A multilingual speech recognition model fine-tuned based on Whisper-base, supporting International Phonetic Alphabet (IPA) output

Snowflake Arctic Embed M v2.0 is a multilingual sentence embedding model focused on sentence similarity tasks, supporting over 50 languages.

AIM-v2 is an image encoder model based on the timm library, suitable for image feature extraction tasks.

VesselFM is a foundation model for universal 3D vascular segmentation in any imaging domain.

BLIP is a unified vision-language pretraining framework, excelling in tasks like image caption generation and visual question answering, with performance enhanced by innovative data filtering mechanisms

ZoeDepth is a depth estimation model fine-tuned on NYU and KITTI datasets, capable of estimating depth values in actual metric units.

ZoeDepth is a model for monocular depth estimation, specifically fine-tuned on the NYU dataset, capable of zero-shot transfer and metric depth estimation.

This is an open-source model under the Apache-2.0 license. Specific information needs to be supplemented.

This is a model released under the MIT license, with specific details currently unknown.

DPT model based on SwinV2 backbone network for monocular depth estimation, trained on 1.4 million images

DPT model based on SwinV2 backbone network for monocular depth estimation, trained on 1.4 million images.

Monocular depth estimation model based on BEiT backbone network, capable of inferring detailed depth information from a single image

A monocular depth estimation model based on Vision Transformer (ViT), trained on 1.4 million images

A scientific term recognition model based on SciBERT, supporting NER-enhanced topic modeling

This model is released under the MIT license, with specific details currently unknown.

InfoXLM is a cross-lingual pre-training framework based on information theory, designed to enhance cross-lingual representation learning by maximizing mutual information between different languages.

GigaBERT-v4 is a model further pretrained on code-mixed data based on GigaBERT-v3, demonstrating improved zero-shot transfer performance from English to Arabic in information extraction (IE) tasks.

GigaBERT-v3 is a bilingual BERT model customized for English and Arabic, pre-trained on a large-scale corpus, and excels in information extraction tasks.

mDeBERTa is the multilingual version of DeBERTa, employing ELECTRA-style pretraining and gradient-disentangled embedding sharing technology, demonstrating excellent performance in cross-lingual tasks like XNLI

A large-scale summarization model based on the BART architecture, fine-tuned specifically on the xsum dataset, excelling at generating concise news summaries.

InfoXLM is a cross-lingual pre-training framework based on information theory, designed to enhance model performance by maximizing mutual information in cross-lingual tasks.

This is a binary classification model based on the XLM-Roberta-base architecture, capable of classifying sentences as facts or opinions. It supports English and Greek, and features zero-shot learning capability.

Open-source model based on Apache-2.0 license (specific information unavailable)