LAION

PROJECTS

DATASETS

LAION-400M

image/text

Status: Released

Formerly known as crawling@home (C@H), an openly accessible 400M image-text-pair dataset.

LAION5B

image/text

Status: Released

A dataset consisting of 5.85 billion CLIP-filtered image-text pairs, featuring several nearest neighbor indices, an improved web-interface for exploration and subset generation, and detection scores for watermark, NSFW, and toxic content detection.

Laion-coco

image/text

Status: Released

600M captions generated using BLIP from Laion2B-en.

Laion translated

image/text

Status: Released

3B translated samples from Laion5B.

Clip H/14

image/text

Status: Released

The largest open source clip.

LAION5B High-Res

image/text

Status: Released

A subset of the LAION5B database, with high resolution images over 1024x1024, containing 170 million samples.

LAION Aesthetics

image/text

Status: Released

A subset of LAION5B that has been estimated by a model trained on top of clip embeddings to contain only aestheticly pleasing images.

LAION-3D

3d/image/text

Status: Started

An effort to create a large-scale dataset consisting of 3D models and descriptor pairs.

Audio Dataset

text/audio

Status: Started

An audio dataset for training CLAP and other models, containing a raw and processed dataset, the latter containing .flac files with captions, labels, and other metadata.

Watermark Detection

image/text

Contrastive

Status: Released

A repository containing datasets to train a watermark classifier.

LAION-BVD

video/text

Status: Released

A large-scale open video dataset for multimodal learning, containing 1.3B platform-specific video URLs collected from CommonCrawl, 80M downloaded videos with a total duration of 10 million hours, 55 million captioned video clips, and 300 million captioned video frames.

MODELS

Openclip

image/text

Contrastive

Status: released

An open source implementation of OpenAI's CLIP (Contrastive Language-Image Pre-training).

DALLE-2 Prior/Decoder

image/text

Generative

Status: Started

An implementation of DALL-E 2, OpenAI's text-to-image synthesis neural network, in Pytorch.

ClipCap

image/text

Generative

Status: Released

Generate text from embedding, using pretrained encoder and language models.

CLAP

audio/text

Contrastive

Status: Started

A Contrastive Language-Audio Pretraining model, like CLIP, for audio.

Video CLIP

video/text

Contrastive

Status: Planning

A contrastive language pretraining model for videos.

Multilingual-CLIP

image/text

Contrastive

Status: In progress

An implementation of OpenAI's CLIP text encoders for any language.

NSFW Detection

image/text

Contrastive

Status: Released

A detector for not safe for work content within images using CLIP.

Electric Sheep

image/text/audio/video

Contrastive/Generative

Status: Started

Train Contrastive and generative models on all modalities.