TAPNext++ (ONNX single graph + PyTorch) for Perspectra

Redistribution of an existing open model, converted for the Perspectra Blender add-on. No retraining.

  • Original work: Google DeepMind, TAPNext++ (https://github.com/google-deepmind/tapnet)
  • Licence: APACHE-2.0 — the same as the original weights. All credit to the original authors.
  • Paper: Zholus et al., TAPNext: Tracking Any Point (TAP) as Next Token Prediction, ICCV 2025
  • tapnextpp_single.onnx: one fp32 graph, 256 x 256 RGB in [-1, 1], 256 queries, recurrent caches as inputs / outputs. tapnextpp.pt: the PyTorch checkpoint used on the GPU.
File Size SHA-256
tapnextpp_single.onnx 979 MB 7bdb8441b274e8fe…
tapnextpp.pt 2,532 MB cb96a43444ccb4fb…
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support