Credits

The footage, models and software that ClipRecall is built on. This page is a notice and a thank you.

Footage

Video: "Multiview Extended Video with Activities" (MEVA) dataset by Kitware Inc. and IARPA, licensed under a Creative Commons Attribution 4.0 International License. Clips are shortened and resized.

Source: mevadata.org. Licence: creativecommons.org/licenses/by/4.0.

The people in the footage are actors who agreed to be filmed.

Models

RF-DETR LargeBy Roboflow, under the Apache License 2.0. Built on the DINOv2 backbone by Meta, Apache License 2.0. It finds people and vehicles.github.com/roboflow/rf-detr
SigLIP 2By Google, under the Apache License 2.0. It turns pictures and words into search numbers.huggingface.co/google
ByteTrack methodBy Yifu Zhang and others. We use the version in Roboflow's trackers library, Apache License 2.0. It follows each mark from frame to frame.github.com/roboflow/trackers

Software

FFmpegReads video and cuts clipsffmpeg.org
OpenCVPicture handling. Apache License 2.0opencv.org
PyTorch and Hugging Face TransformersRun the models. BSD style licences and Apache License 2.0pytorch.org
NumPy and SciPyArithmetic. BSD style licencesnumpy.org
Roboflow supervisionBox records. MIT Licensegithub.com/roboflow/supervision
FastAPIThe service. MIT Licensefastapi.tiangolo.com
PostgreSQLThe database. PostgreSQL Licencewww.postgresql.org
Next.js and ReactThe web pages. MIT Licensenextjs.org
three.jsDraws the 3D cameras and devices in the page. MIT Licensethreejs.org
LenisSmooth scrolling. MIT Licensegithub.com/darkroomengineering/lenis
Onest and Geist MonoThe typefaces. SIL Open Font License 1.1fontsource.org

What a credit does not mean

None of the people or organisations named here supports or approves ClipRecall, and none of them is a ClipRecall customer.