## Requirements
In [[Accelerated Learning in the Context of Hand-Object Interaction|my research on joint hand-object pose estimation]], and because my focus is on object-specific adaptation, I find myself in the need of different datasets than the typical use. There are various hand-object interaction datasets out there, but only a few meet my requirements:
1. Provide the ground-truth for the 2D or 3D hand joint poses (21 joints)
2. Provide the ground-truth for the 6D object pose
3. Contain several object categories/shapes; the more the better
4. Provide RGB images
For my current experiments with object-specific adaptation with [[An in-depth view of the MAML algorithm|meta-learning]], I need datasets that contain as many different objects as possible, in order to exploit the advantage of meta-learning methods for few-shot learning, since the goal is to demonstrate the potential increase in the ability of a model to generalize to unseen objects.
## Curated list of datasets
| Dataset | # subjects | # frames | # objects | Nature | Year |
| ---------------------------------------------------------------------------------------------------------------------- | ---------- | -------- | ---------------- | --------------- | ---- |
| [SynthHands](http://occludedhands.com/) | 2 | 63,350 | 7 | Synthetic | 2017 |
| [FPHAD](https://github.com/guiggh/hand_pose_action) | 6 | 100,000 | 4 | Real | 2018 |
| [FHAB](https://guiggh.github.io/publications/first-person-hands/) | 6 | 105,459 | 26 (4 annotated) | Real | 2018 |
| [ObMan](https://hassony2.github.io/obman) | N/A | 150,000 | 8 | Synthetic | 2019 |
| [ContactPose](https://research.fb.com/publications/contactpose-a-dataset-of-grasps-with-object-contact-and-hand-pose/) | 50 | 2.9M | 25 | Hybrid | 2020 |
| [GRAB](https://grab.is.tue.mpg.de/) | 10 | 1.6M | 51 | Mocap | 2020 |
| [HO3D](https://www.tugraz.at/institute/icg/research/team-lepetit/research-projects/hand-object-3d-pose-annotation/) | 10 | 77,558 | 10 | Real | 2021 |
| [H2O](https://taeinkwon.com/projects/h2o/) | 4 | 571,645 | 8 | Real | 2021 |
| [DexYCB](https://dex-ycb.github.io/) | 10 | 582,000 | 20 | Real | 2021 |
| [OakInk](https://oakink.net/) | 12 | 230,000 | 100 | Real | 2022 |
| [HOI4D](https://hoi4d.github.io/) | 9 | 2.4M | 800 | Real egocentric | 2022 |
| [ARCTIC](https://arctic.is.tue.mpg.de/) | 10 | 2.1M | 11 | Real | 2023 |
### Related datasets not meeting the requirements
| Dataset | # subjects | # frames | # objects | Nature | Year | Reason |
| ----------------------------------------------------------------------------------------- | ---------- | -------- | --------- | --------- | ---- | ---------------------------------------------------------- |
| [Dexter+Object](https://handtracker.mpi-inf.mpg.de/projects/RealtimeHO/dexter+object.htm) | 4 | 3,014 | N/A | Real | 2016 | Only finger tip annotations. |
| [HUST Dataset](https://www.handcorpus.org/?p=1596) | 30 | N/A | N/A | Real | 2016 | Cyberglove used, no object annotations, only joint angles. |
| [TUB Dataset](https://www.handcorpus.org/?p=1830) | 17 | N/A | 25 | Real | 2016 | Cyberglove used, no object annotations, only joint angles. |
| [EgoDexter](https://handtracker.mpi-inf.mpg.de/projects/OccludedHands/EgoDexter.htm) | 4 | 1,485 | N/A | Real | 2017 | Only finger tip annotations. |
| [Rendered Hand Pose (RHD)](https://lmb.informatik.uni-freiburg.de/projects/hand3d/) | 20 | 44,700 | N/A | Synthetic | 2017 | No object annotations. |
| [UNIPI Dataset](https://www.handcorpus.org/?p=1855) | 6 | N/A | 21 | Real | 2018 | No object annotations, only joint angles. |
| [Partially Occluded Hands](http://occludedhands.com/) | N/A | 11,840 | 148 | Real | 2018 | No object annotations. |
| [FreiHAND](https://lmb.informatik.uni-freiburg.de/projects/freihand/) | N/A | 130,3960 | N/A | Real | 2019 | No object annotations. |
| [ContactDB](https://contactdb.cc.gatech.edu/) | 50 | 375K | 50 | Real | 2019 | No joint annotations. |
---
 [^1]
[^1]: [Survey on depth and RGB image-based 3D hand shape and pose estimation (Huang et al. 2021)](https://doi.org/10.1016/j.vrih.2021.05.002)