Dense object tracking, the ability to localize specific object points with pixel-level accuracy, is an important computer vision task with numerous downstream applications in robotics. Existing approaches either compute dense keypoint embeddings in a single forward pass, meaning the model is trained to track everything at once, or allocate their full capacity to a sparse predefined set of points, trading generality for accuracy. In this paper we explore a middle ground based on the observation that the number of relevant points at a given time are typically relatively few, e.g. grasp points on a target object. Our main contribution is a novel architecture, inspired by few-shot task adaptation, which allows a sparse-style network to condition on a keypoint embedding that indicates which point to track. Our central finding is that this approach provides the generality of dense-embedding models, while offering accuracy significantly closer to sparse-keypoint approaches. We present results illustrating this capacity vs. accuracy trade-off, and demonstrate the ability to zero-shot transfer to new object instances (within-class) using a real-robot pick-and-place task.


    Access

    Download


    Export, share and cite



    Title :

    Few-Shot Keypoint Detection as Task Adaptation via Latent Embeddings


    Contributors:
    Vecerik, M (author) / Kay, J (author) / Hadsell, R (author) / Agapito, L (author) / Scholz, J (author)

    Publication date :

    2022-07-12


    Remarks:

    In: Proceedings - IEEE International Conference on Robotics and Automation. (pp. pp. 1251-1257). IEEE: Philadelphia, PA, USA. (2022)


    Type of media :

    Paper


    Type of material :

    Electronic Resource


    Language :

    English



    Classification :

    DDC:    629



    MONOCULAR 2D SEMANTIC KEYPOINT DETECTION AND TRACKING

    CHEN HAOFENG / ARJUN BHARGAVA / RARES ANDREI AMBRUS et al. | European Patent Office | 2023

    Free access

    Keypoint based action localization

    KADAV ASIM / LAI FARLEY / GRAF HANS PETER et al. | European Patent Office | 2025

    Free access

    KEYPOINT BASED ACTION LOCALIZATION

    KADAV ASIM / LAI FARLEY / GRAF HANS PETER et al. | European Patent Office | 2022

    Free access


    Test-Time Adaptation for Keypoint-Based Spacecraft Pose Estimation Based on Predicted-View Synthesis

    Bravo, Juan Ignacio Ignacio / Garcia-Martin, Alvaro / Bescos, Jesus et al. | IEEE | 2024