# Cross-Architecture Auxiliary Feature Space Translation for Efficient Few-Shot Personalized Object Detection Source: [Cross-Architecture Auxiliary Feature Space Translation for Efficient Few-Shot Personalized Object Detection]() Collection index: [llms.txt]() Data updated: 2025-10-26T02:59:03.375Z Authors: Francesco Barbato, Umberto Michieli, Jijoong Moon, Pietro Zanuttigh, Mete Ozay Published: 2024-07-01T11:33:53.000Z ## Abstract Recent years have seen object detection robotic systems deployed in several personal devices (e.g., home robots and appliances). This has highlighted a challenge in their design, i.e., they cannot efficiently update their knowledge to distinguish between general classes and user-specific instances (e.g., a dog vs. user's dog). We refer to this challenging task as Instance-level Personalized Object Detection (IPOD). The personalization task requires many samples for model tuning and optimization in a centralized server, raising privacy concerns. An alternative is provided by approaches based on recent large-scale Foundation Models, but their compute costs preclude on-device applications. In our work we tackle both problems at the same time, designing a Few-Shot IPOD strategy called AuXFT. We introduce a conditional coarse-to-fine few-shot learner to refine the coarse predictions made by an efficient object detector, showing that using an off-the-shelf model leads to poor personalization due to neural collapse. Therefore, we introduce a Translator block that generates an auxiliary feature space where features generated by a self-supervised model (e.g., DINOv2) are distilled without impacting the performance of the detector. We validate AuXFT on three publicly available datasets and one in-house benchmark designed for the IPOD task, achieving remarkable gains in all considered scenarios with excellent time-complexity trade-off: AuXFT reaches a performance of 80% its upper bound at just 32% of the inference time, 13% of VRAM and 19% of the model size. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Personalized Object Detection on CORe50 1-shot](): Personalized Object Detection; CORe50 1-shot; mAP: 58.8. - [Personalized Object Detection on CORe50 5-shot](): Personalized Object Detection; CORe50 5-shot; mAP: 69.3. - [Personalized Object Detection on POD 5-shot (mAP)](): Personalized Object Detection; POD 5-shot; mAP: 38.8. - [Personalized Object Detection on iCubWorld 5-shot](): Personalized Object Detection; iCubWorld 5-shot; mAP: 74.5. - [Personalized Object Detection on PerSeg 1-shot](): Personalized Object Detection; PerSeg 1-shot; mAP: 48.8. - [Personalized Object Detection on iCubWorld 1-shot](): Personalized Object Detection; iCubWorld 1-shot; mAP: 55. ## Original sources and implementation [Original paper]() [PDF]()