# facebookresearch/perception_models

State-of-the-art Image & Video CLIP, Multimodal Large Language Models, and More!

Repository: https://github.com/facebookresearch/perception_models
Canonical: https://ross.abutalabs.com/products/perception_models
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Last push: 2026-04-13T21:42:56+00:00

## Health v2 (maintenance only)
Score: 54/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 77, release rhythm 35, longevity 36
- inputs: {"age_days": 504, "days_push": 142, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2353, forks 160 (observed 2026-08-28T04:06:40.487483+00:00)

## What it is
Meta's Perception Models repository hosting state-of-the-art image, video, and audio encoders (Perception Encoder, PE) and a multimodal language model (Perception Language Model, PLM). It provides pretrained checkpoints, evaluation benchmarks, and integrations with Hugging Face transformers and timm.

## Use cases
- encode images and videos into embeddings for retrieval
- run zero-shot image and video classification with CLIP-style models
- build video question answering with a multimodal language model
- evaluate multimodal models on PLM-VideoBench
- extract audio-visual embeddings for cross-modal search
- fine-tune vision-language models on custom video data

## When to choose
- you need state-of-the-art visual or audiovisual embeddings for retrieval or classification
- you want an open multimodal LLM for video understanding tasks
- you want models already integrated into Hugging Face transformers or timm

## When to avoid
- you need lightweight CPU-only inference without GPU resources
- you need a production application rather than research models and checkpoints
- you need text-only language modeling without vision or audio

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, computer-vision, image-processing, video-processing, audio-processing, llm-inference, transformers
- domain: computer-vision, deep-learning, machine-learning, artificial-intelligence
- platform: python, cross-platform
- tags: clip, perception-encoder, perception-language-model, multimodal, embeddings, video-understanding, audiovisual, pytorch, model-checkpoints, facebook-research, video, audio, natural-language-processing, gpu

## Member repositories
- facebookresearch/perception_models (main) score 54

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:40.487483+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:36:29.473148+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/perception_models (fetched 2026-08-28T04:06:40.487483+00:00, sha c7791a380044)
- Data as of 2026-08-30T08:39:29.467469+00:00.