A latent text-to-image diffusion model capable of generating photo-realistic images given any text input

The stable-diffusion model is a text-to-image diffusion model that can generate highly realistic images based on textual input. It utilizes latent variables and diffusion processes to achieve this.

Practical face restoration algorithm for *old photos* or *AI-generated faces*

GFPGAN is a practical face restoration algorithm designed for improving the quality of old photos or AI-generated faces. It uses a generator network to enhance facial details by upscaling low-resolution images and refining facial features. The model's objective is to generate high-quality, realistic face images that restore and improve the appearance of the input images.

Bootstrapping Language-Image Pre-training

The Bootstrapping Language-Image Pre-training (BLIP) model is a technique that integrates text and image data to improve language understanding and generation tasks. It uses a pre-training stage where training data is combined from both images and text to learn joint representations. These joint representations are then used to fine-tune the model on specific downstream tasks, such as image captioning or text generation. BLIP has been shown to surpass state-of-the-art results on various tasks and can be applied to a wide range of applications that involve both text and image data.

Platform did not provide a description for this model.

The model wav2vec2-large-xlsr-53-english is an automatic speech recognition (ASR) model designed to convert spoken language into written text. It is trained using the wav2vec2 architecture and the Cross-lingual Speaker Representations (XLSR) method. The model is specifically trained for the English language and is capable of accurately transcribing speech for various applications such as transcription services, voice assistants, and voice command recognition systems.

Return CLIP features for the clip-vit-large-patch14 model

The clip-features model is a model that utilizes the clip-vit-large-patch14 architecture to extract features from text and images. It takes in an image and text as input and returns the corresponding CLIP features. These features can then be used for various tasks such as image classification, object detection, and image generation. The model is designed to provide a compact representation of the input that captures both visual and textual information, allowing for cross-modal understanding and analysis.

Pretrained model on English language using a masked language modeling (MLM) objective. It was introduced in
this paper and first released in
this repository. This model is uncased: it does not make a difference
between english and English.

Disclaimer: The team releasing BERT did not write a model card for this model so this model card has been written by
the Hugging Face team.

BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it
was pretrained on the raw texts only, with no humans labeling them in any way (which is why it can use lots of
publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it
was pretrained with two objectives:

This way, the model learns an inner representation of the English language that can then be used to extract features
useful for downstream tasks: if you have a dataset of labeled sentences, for instance, you can train a standard
classifier using the features produced by the BERT model as inputs.

BERT has originally been released in base and large variations, for cased and uncased input text. The uncased models also strips out an accent markers.Chinese and multilingual uncased and cased versions followed shortly after.Modified preprocessing with whole word masking has replaced subpiece masking in a following work, with the release of two models.Other 24 smaller models are released afterward.  

The detailed release history can be found on the google-research/bert readme on github.

You can use the raw model for either masked language modeling or next sentence prediction, but it's mostly intended to
be fine-tuned on a downstream task. See the model hub to look for
fine-tuned versions of a task that interests you.

Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked)
to make decisions, such as sequence classification, token classification or question answering. For tasks such as text
generation you should look at model like GPT2.

You can use this model directly with a pipeline for masked language modeling:

Here is how to use this model to get the features of a given text in PyTorch:

and in TensorFlow:

Even if the training data used for this model could be characterized as fairly neutral, this model can have biased
predictions:

This bias will also affect all fine-tuned versions of this model.

The BERT model was pretrained on BookCorpus, a dataset consisting of 11,038
unpublished books and English Wikipedia (excluding lists, tables and
headers).

The texts are lowercased and tokenized using WordPiece and a vocabulary size of 30,000. The inputs of the model are
then of the form:

With probability 0.5, sentence A and sentence B correspond to two consecutive sentences in the original corpus, and in
the other cases, it's another random sentence in the corpus. Note that what is considered a sentence here is a
consecutive span of text usually longer than a single sentence. The only constrain is that the result with the two
"sentences" has a combined length of less than 512 tokens.

The details of the masking procedure for each sentence are the following:

The model was trained on 4 cloud TPUs in Pod configuration (16 TPU chips total) for one million steps with a batch size
of 256. The sequence length was limited to 128 tokens for 90% of the steps and 512 for the remaining 10%. The optimizer
used is Adam with a learning rate of 1e-4, β1=0.9\beta_{1} = 0.9β1​=0.9 and β2=0.999\beta_{2} = 0.999β2​=0.999, a weight decay of 0.01,
learning rate warmup for 10,000 steps and linear decay of the learning rate after.

When fine-tuned on downstream tasks, this model achieves the following results:

Glue test results:



BERT-base-uncased is a pretrained language model that has been trained on a large corpus of English text. It uses a masked language modeling objective to learn an inner representation of the English language. This pretrained model can be used for tasks such as sequence classification, token classification, and question answering. It has been trained on the BookCorpus dataset and English Wikipedia using a vocabulary size of 30,000. The model has been trained on 4 cloud TPUs with a batch size of 256 and uses the Adam optimizer. When fine-tuned on downstream tasks, BERT-base-uncased achieves good performance on tasks like sentiment analysis and text classification. However, it is important to note that the model may have biased predictions, and this bias can also affect all fine-tuned versions of the model.

A text-to-image generative AI model that creates beautiful images

The sdxl model is a text-to-image generative AI model that can create high-resolution images with a size of 1024x1024 pixels. It takes text descriptions as input and generates corresponding images as output. The model aims to create beautiful and realistic images based on the provided textual prompts.

Real-ESRGAN with optional face correction and adjustable upscale

Real-ESRGAN is a model that can enhance the resolution of low-resolution images, providing a higher level of detail and clarity. It uses a technique called enhanced super-resolution generative adversarial networks (ESRGAN) to achieve this. The model also offers the ability to correct faces in the images and adjust the level of upscaling for fine-grained control.

Generate detailed images from scribbled drawings

The controlnet-scribble model is a text-to-image model that can generate detailed images from scribbled drawings. It uses a neural network architecture called ControlGAN, which incorporates a control module to guide the generation process based on textual descriptions. This model can be useful for tasks such as image synthesis, where detailed images need to be generated based on simple input sketches or descriptions.

Robust face restoration algorithm for old photos / AI-generated faces

Codeformer is a robust face restoration algorithm that can restore old photos and generate AI-generated faces. It uses advanced AI techniques to analyze and process images, making them look as good as new. The model is designed to be highly effective in restoring and enhancing the quality of old and degraded photographs, as well as generating highly realistic AI-generated faces. This algorithm could be incredibly useful for tasks like photo restoration, face synthesis, and other image-to-image applications.

The image-tagger is a model that can generate textual descriptions of images. It takes an input image and provides a descriptive text output, highlighting the key features and objects present in the image. This model can be useful in applications that involve image recognition and understanding, such as captioning images for visually impaired people or indexing and searching images based on their content.

Models