Update Model Card with complete checkpoint listing, license, and usage examples

README.md

@@ -1,22 +1,197 @@
 ---
 license: apache-2.0
 tags:
-- vision
-- tracking
+  - vision
+  - tracking
+  - point-tracking
+  - video
+  - pytorch
+  - jax
+library_name: transformers
+arxiv:
+  - "2211.03726"
+  - "2306.08637"
+  - "2402.00847"
+  - "2504.05579"
+  - "2604.10582"
 ---
 
-# TAPNet
+# Tracking Any Point (TAP) — Model Checkpoints
 
-This repository contains the checkpoints of several point tracking models developed by DeepMind for point tracking.
+Official pre-trained checkpoints for the **Tracking Any Point (TAP)** family of models from Google DeepMind & Google Research.
 
-🔗 **Code**: [https://github.com/google-deepmind/tapnet](https://github.com/google-deepmind/tapnet)  
+[[GitHub](https://github.com/google-deepmind/tapnet)] [[TAP-Vid](https://tapvid.github.io/)] [[TAPIR](https://deepmind-tapir.github.io/)] [[BootsTAP](https://bootstap.github.io/)] [[TAPNext](https://tap-next.github.io/)] [[TAPNext++](https://tap-next-plus-plus.github.io/)]
 
-## Included Models
+## Model Overview
 
-[**TAPIR**](https://deepmind-tapir.github.io/) – A fast and accurate point tracker for continuous point trajectories in space-time, presented in the paper [**TAPIR: Tracking Any Point with per-frame Initialization and temporal Refinement**](https://huggingface.co/papers/2306.08637).
+The TAP model family addresses the task of **tracking any point on any solid physical surface** across video frames. Given a query point on one frame, the model predicts where that point moves to on every other frame, along with an occlusion/visibility estimate.
 
-[**BootsTAPIR**](https://bootstap.github.io/) – A bootstrapped variant of TAPIR that improves robustness and stability across long videos via self-supervised refinement, presented in the paper [**BootsTAP: Bootstrapped Training for Tracking-Any-Point**](https://huggingface.co/papers/2402.00847).
-  
-[**TAPNext**](https://tap-next.github.io/) – A new generative approach that frames point tracking as next-token prediction, enabling semi-dense, accurate, and temporally coherent tracking across challenging videos, presented in the paper [**TAPNext: Tracking Any Point (TAP) as Next Token Prediction**](https://huggingface.co/papers/2504.05579).
+| Model | Key Contribution |
+|:--|:--|
+| **TAP-Net** | Baseline model introduced with the TAP-Vid benchmark |
+| **TAPIR** | Two-stage matching + refinement; large accuracy gains over TAP-Net |
+| **BootsTAPIR** | TAPIR + bootstrapped self-training on unlabeled video |
+| **TAPNext / BootsTAPNext** | Next-token-prediction formulation; fastest and simplest tracker |
+| **TAPNext++** | Fine-tuned BootsTAPNext with 40× longer stable tracking, occlusion handling, and re-detection |
+| **TRAJAN** | Point trajectory autoencoder for motion representation learning |
 
-These models provide state-of-the-art performance for tracking arbitrary points in videos and support research and applications in robotics, perception, and video generation.
+## Checkpoints
+
+All checkpoints are released under the **Apache License 2.0**.
+
+### TAPNext++ (latest)
+
+| File | Resolution | Framework | DAVIS First (AJ) | Kinetics First (AJ) | RoboTAP First (AJ) |
+|:--|:--:|:--:|:--:|:--:|:--:|
+| `tapnextpp_ckpt.pt` | 256×256 | PyTorch | 65.6% | 53.9% | 61.1% |
+| `tapnextpp_512.ckpt` | 512×512 | PyTorch | 67.0% | — | — |
+
+### TAPNext / BootsTAPNext
+
+| File | Resolution | Framework | DAVIS First (AJ) |
+|:--|:--:|:--:|:--:|
+| `bootstapnext.pth` | 256×256 | PyTorch | 65.25% |
+| `bootstapnext_ckpt.npz` | 256×256 | Jax | 65.25% |
+
+### TAPIR & BootsTAPIR
+
+| File | Mode | Resolution | Framework | DAVIS First (AJ) |
+|:--|:--:|:--:|:--:|:--:|
+| `tapir_checkpoint_panning.pt` | Offline | 256×256 | PyTorch | 58.5% |
+| `tapir_checkpoint_panning.npy` | Offline | 256×256 | Jax | 58.5% |
+| `causal_tapir_checkpoint.npy` | Online | 256×256 | Jax | 56.2% |
+| `bootstapir_checkpoint_v2.pt` | Offline | 256×256 + 512×512 | PyTorch | 62.4% |
+| `bootstapir_checkpoint_v2.npy` | Offline | 256×256 + 512×512 | Jax | 62.4% |
+| `causal_bootstapir_checkpoint.pt` | Online | 256×256 + 512×512 | PyTorch | 59.7% |
+| `causal_bootstapir_checkpoint.npy` | Online | 256×256 + 512×512 | Jax | 59.7% |
+
+### Other
+
+| File | Model | Framework |
+|:--|:--|:--:|
+| `checkpoint.npy` | TAP-Net (baseline) | Jax |
+| `track_autoencoder_ckpt.npz` | TRAJAN | Jax |
+
+## Quick Start
+
+### TAPNext++ (PyTorch)
+
+```python
+import torch
+from tapnet.torch import tapnextpp
+
+# Load model
+model = tapnextpp.TAPNextPlusPlus()
+checkpoint = torch.load("tapnextpp_ckpt.pt", map_location="cpu")
+model.load_state_dict(checkpoint)
+model.eval()
+```
+
+See the [TAPNext++ Colab](https://colab.research.google.com/github/deepmind/tapnet/blob/main/colabs/torch_tapnextpp_demo.ipynb) for a complete demo.
+
+### BootsTAPIR (PyTorch)
+
+```python
+import torch
+from tapnet.torch import tapir_model
+
+# Load model
+model = tapir_model.ParameterizedTAPIR()
+checkpoint = torch.load("bootstapir_checkpoint_v2.pt", map_location="cpu")
+model.load_state_dict(checkpoint)
+model.eval()
+```
+
+See the [PyTorch TAPIR Colab](https://colab.research.google.com/github/deepmind/tapnet/blob/master/colabs/torch_tapir_demo.ipynb) for a complete demo.
+
+## Installation
+
+```bash
+pip install tapnet
+# or with PyTorch support
+pip install "tapnet[torch]"
+```
+
+## Citation
+
+If you use these models, please cite the relevant papers:
+
+<details>
+<summary>TAP-Vid (NeurIPS 2022)</summary>
+
+```bibtex
+@article{doersch2022tap,
+  title={{TAP}-Vid: A Benchmark for Tracking Any Point in a Video},
+  author={Doersch, Carl and Gupta, Ankush and Markeeva, Larisa and Recasens, Adria and Smaira, Lucas and Aytar, Yusuf and Carreira, Joao and Zisserman, Andrew and Yang, Yi},
+  journal={Advances in Neural Information Processing Systems},
+  volume={35},
+  pages={13610--13626},
+  year={2022}
+}
+```
+</details>
+
+<details>
+<summary>TAPIR (ICCV 2023)</summary>
+
+```bibtex
+@inproceedings{doersch2023tapir,
+  title={{TAPIR}: Tracking any point with per-frame initialization and temporal refinement},
+  author={Doersch, Carl and Yang, Yi and Vecerik, Mel and Gokay, Dilara and Gupta, Ankush and Aytar, Yusuf and Carreira, Joao and Zisserman, Andrew},
+  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
+  pages={10061--10072},
+  year={2023}
+}
+```
+</details>
+
+<details>
+<summary>BootsTAP (ACCV 2024)</summary>
+
+```bibtex
+@article{doersch2024bootstap,
+  title={{BootsTAP}: Bootstrapped Training for Tracking-Any-Point},
+  author={Doersch, Carl and Luc, Pauline and Yang, Yi and Gokay, Dilara and Koppula, Skanda and Gupta, Ankush and Heyward, Joseph and Rocco, Ignacio and Goroshin, Ross and Carreira, Jo{\~a}o and Zisserman, Andrew},
+  journal={Asian Conference on Computer Vision},
+  year={2024}
+}
+```
+</details>
+
+<details>
+<summary>TAPNext (2025)</summary>
+
+```bibtex
+@article{zholus2025tapnext,
+  title={TAPNext: Tracking Any Point (TAP) as Next Token Prediction},
+  author={Zholus, Artem and Doersch, Carl and Yang, Yi and Koppula, Skanda and Patraucean, Viorica and He, Xu Owen and Rocco, Ignacio and Sajjadi, Mehdi S. M. and Chandar, Sarath and Goroshin, Ross},
+  journal={arXiv preprint arXiv:2504.05579},
+  year={2025}
+}
+```
+</details>
+
+<details>
+<summary>TAPNext++ (CVPR 2026 Findings)</summary>
+
+```bibtex
+@InProceedings{Jung_2026_CVPR,
+  author    = {Jung, Sebastian and Zholus, Artem and Sundermeyer, Martin and Doersch, Carl
+               and Goroshin, Ross and Tan, David Joseph and Chandar, Sarath
+               and Triebel, Rudolph and Tombari, Federico},
+  title     = {TAPNext++: What's Next for Tracking Any Point (TAP)?},
+  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Findings},
+  month     = {June},
+  year      = {2026},
+  pages     = {8429-8438}
+}
+```
+</details>
+
+## License
+
+Copyright 2022-2026 Google LLC
+
+All software and model checkpoints in this repository are licensed under the [Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0).
+
+This is not an official Google product.