LAPA
Collection
LAPA (Look Around and Pay Attention): multi-camera point tracking. Contains pretrained models for TAPVid-3D-MC and PointOdyssey-MC and LAPA-Joint • 4 items • Updated
LAPA
Look Around and Pay Attention — multi-camera point tracking with transformers.
Paper: arXiv:2512.04213
Code: GitHub
Collection: LAPA on Hugging Face
Pretrained models
| Model | Dataset | Hub |
|---|---|---|
| LAPA TAPVid-3D-MC | TAPVid-3D-MC | bishoygaloaa/LAPA-TAPVid-3D-MC |
| LAPA PointOdyssey-MC | PointOdyssey-MC | bishoygaloaa/LAPA-PointOdyssey-MC |
Load
import torch
from huggingface_hub import hf_hub_download
from lapa.models.lapa import LAPA
# TAPVid-3D-MC
path = hf_hub_download("bishoygaloaa/LAPA-TAPVid-3D-MC", "lapa.pt")
# PointOdyssey-MC (when available)
# path = hf_hub_download("bishoygaloaa/LAPA-PointOdyssey-MC", "lapa.pt")
ckpt = torch.load(path, map_location="cpu")
model = LAPA()
model.load_state_dict(ckpt["model"])
model.eval()
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support