Hand-Object Interaction & Affordance
Robust 3D Hand-Object Pose and Interaction Estimation Under Upstream 2D Segmentation and Localization Noise
Generated automatically from the limitations stated in 9 papers (ICLR, CVPR, ECCV), listed under Evidence. It is not a paper, and it does not come from papers submitted to CSPaper.
The problem
Current hand-object interaction and 3D hand pose estimation models rely heavily on upstream 2D bounding box crops, hand segmentation masks, or pre-extracted interaction regions. In real-world egocentric scenes, upstream 2D segmenters and detectors degrade severely due to object occlusions, motion blur, and cluttered backgrounds, causing downstream 3D estimation to fail or hallucinate. Because models are trained and benchmarked assuming near-ideal 2D crops or masks, existing pipelines cannot operate reliably in autonomous, in-the-wild video streams.
Why it matters
End-to-end hand-object pose and affordance estimation directly from uncurated egocentric video streams without requiring clean manual prompts, high-accuracy 2D segmentation masks, or heuristic threshold tuning.
Ways to approach it
Prior-work checks are free with an account. Results someone already ran are shown to everyone.
- 1
Benchmark degradation by evaluating state-of-the-art 3D hand/HOI estimators under controlled perturbations of upstream 2D bounding boxes and segmentation masks (dilation, erosion, boundary noise, and synthetic occlusion) on datasets like VISOR and HO-3D to quantify error propagation.
- 2
Train end-to-end full-image HOI architectures that integrate soft, multi-scale hand attention mechanisms instead of hard bounding box crops or binary masks, measuring 3D joint error under noisy detector proposals.
- 3
Design occlusion-aware feature aggregation that conditions joint estimation on temporal optical flow and kinematic priors rather than local mask resamplers, measuring pose accuracy during severe object occlusion phases (e.g., manipulation, folding).
Have a different approach?
Describe how you would tackle this problem and we'll look for papers that already do it. Free; your text stays private.
Why it might fail
If upstream foundation models (e.g., future iterations of SAM or open-world video object segmenters) become sufficiently robust to cluttered egocentric hands and severe object occlusions that upstream segmentation noise ceases to be a failure mode in practice.
Evidence
Each paper's own statement of the limitation, verbatim.
- Interaction-aware Representation Modeling With Co-Occurrence Consistency for Egocentric Hand-Object ParsingICLR 2026
Performance is sensitive to the hand pixel count threshold hyperparameter (tau) in the CoCo loss, risking false positive hand detections if too low or missing partially visible hands if too high
- GenHOI: Towards Object-Consistent Hand-Object Interaction with Temporally Balanced and Spatially Selective Object InjectionCVPR 2026
Requires binary mask videos defining hand-object interaction regions.
- Enhancing Hands in 3D Whole-Body Pose Estimation with Conditional Hands ModulatorCVPR 2026
Performance depends on the quality of upstream hand localization and 2D bounding box crops.
- EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric VideoICLR 2026
Hand pose annotations rely on on-device visual tracking algorithms and can fail or lose confidence during severe object occlusions (such as folding laundry) or high-speed hand movements.
- EgoDTM: Towards 3D-Aware Egocentric Video-Language PretrainingNeurIPS 2025
Hand segmentation quality is substantially below supervised upper bound (61.01 vs 90.87 hand-AP on VISOR) due to noisy HOI detection prompts and SAM2 struggles in cluttered egocentric backgrounds
- Controlling the World by Sleight of HandECCV 2024
Performance degrades with noisy or low-detail hand masks; more detailed segmentation masks produce finer-grained outputs
- HandR2N2: Iterative 3D Hand Pose Estimation Using a Residual Recurrent Neural NetworkICCV 2023
Performance degrades when hand parts are severely occluded because the resampler cannot sample informative points near occluded joints
- Point-to-Point Regression PointNet for 3D Hand Pose EstimationECCV 2018
Requires hand segmentation (via a separate hourglass network on NYU/ICVL) and PCA pose prior fitting
Show all 9 papers
- Learning to Predict Gaze in Egocentric VideoICCV 2013
Performance is sensitive to the accuracy of hand segmentation and configuration classification, which remains a non-trivial task.
Nearest existing work
- 3D Interacting Hand Pose Estimation by Hand De-Occlusion and RemovalECCV 2022
- Real-Time Hand Tracking Under Occlusion From an Egocentric RGB-D SensorICCV 2017
- Benchmarks and Challenges in Pose Estimation for Egocentric Hand Interactions with ObjectsECCV 2024
- 3D Hand Pose Estimation in Everyday Egocentric ImagesECCV 2024
- Towards A Richer 2D Understanding of Hands at ScaleNeurIPS 2023
- "Fine-Grained Egocentric Hand-Object Segmentation: Dataset, Model, and Applications"ECCV 2022
- Lending A Hand: Detecting Hands and Recognizing Activities in Complex Egocentric InteractionsICCV 2015
- Semi-Supervised 3D Hand-Object Poses Estimation With Interactions in TimeCVPR 2021
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from VideoCVPR 2024
- End-to-End Detection and Pose Estimation of Two Interacting HandsICCV 2021
- MEgoHand: Multimodal Egocentric Hand-Object Interaction Motion GenerationNeurIPS 2025
- Glove2Hand: Synthesizing Natural Hand-Object Interaction from Multi-Modal Sensing GlovesCVPR 2026
- SHOW3D: Capturing Scenes of 3D Hands and Objects in the WildCVPR 2026
- Transformer-Based Unified Recognition of Two Hands Manipulating ObjectsCVPR 2023
- UniHOPE: A Unified Approach for Hand-Only and Hand-Object Pose EstimationCVPR 2025
Related open problems
Hand-Object Interaction & Affordance
Robust Hand-Object Interaction Learning Beyond Cascaded Parametric Pose Estimators
Downstream hand-object interaction (HOI) and affordance models are universally bottlenecked by a strict reliance on upstream off-the-shelf 3D hand pose estimators (e.g., FrankMocap, HaMeR, MANO parameterizations) and bounding-box detectors. When these upstream tools encounter severe object occlusions, fast manipulation dynamics, or egocentric perspective distortion, their estimation errors cascade directly into downstream HOI training, causing catastrophic mis-projections and synthetic contact artifacts. Because downstream methods require clean parametric 3D meshes as preconditions, massive repositories of in-the-wild manipulation videos cannot be utilized without fragile filtering heuristics that discard over half the data.
Hand-Object Interaction & Affordance
Benchmarking Hand-Object Interaction Models Across Diverse Hand Morphologies and Embodiments
Existing hand representation and hand-object interaction models are predominantly evaluated on small demographic cohorts or fixed kinematic setups (e.g., standard anthropomorphic models like MANO or fixed robotic hands like ShadowHand). Because existing pipelines implicitly bind kinematics, palm-to-finger ratios, or subject identities into their representations, their performance on unseen human hand proportions and non-standard robotic end-effectors remains untested. As a result, it is unknown whether current interaction and affordance estimators degrade gracefully or fail catastrophically when applied outside their narrow training morphologies.
Hand-Object Interaction & Affordance
Fine-Grained 3D Hand Affordance Grounding and Generation for Small-Part Articulated Interactions
Current affordance learning pipelines rely on self-training loops, video extraction, or upstream generative HOI models that fail when contact targets are small functional components such as bottle caps, zippers, or rotary knobs. Because generated hand poses easily drift to adjacent object surfaces, models exhibit catastrophic failure on precision affordances (e.g., 0% accuracy on pull or sub-50% on twist), propagating noisy pseudo-labels across unobserved geometries. Downstream manipulation systems are consequently blocked from executing precision tasks like uncapping, pulling sliders, or fine bimanual manipulation.
Hand-Object Interaction & Affordance
Hand-Object Affordance Estimation Beyond Clean Pre-Scanned Scenes and Predefined 3D Grids
Current methods for 3D hand-object interaction and affordance estimation depend strictly on clean geometric preconditions: high-fidelity 3D scans, pre-detected object instances, static overhead views, or tightly predefined 3D bounding grids. In practical manipulation and augmented reality scenarios, environments are unsegmented, viewpoints are dynamic, and point clouds are noisy and incomplete. Relying on manually tuned grid volumes and pre-extracted meshes prevents existing models from operating directly on raw, uncurated sensor streams.