PicoSAM3: Real-Time In-Sensor Region-of-Interest Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bonazzi, Pietro, Farronato, Nicola, Zihlmann, Stefan, Qin, Haotong, Magno, Michele
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912963329261568
author Bonazzi, Pietro
Farronato, Nicola
Zihlmann, Stefan
Qin, Haotong
Magno, Michele
author_facet Bonazzi, Pietro
Farronato, Nicola
Zihlmann, Stefan
Qin, Haotong
Magno, Michele
contents Real-time, on-device segmentation is critical for latency-sensitive and privacy-aware applications such as smart glasses and Internet-of-Things devices. We introduce PicoSAM3, a lightweight promptable visual segmentation model optimized for edge and in-sensor execution, including deployment on the Sony IMX500 vision sensor. PicoSAM3 has 1.3 M parameters and combines a dense CNN architecture with region of interest prompt encoding, Efficient Channel Attention, and knowledge distillation from SAM2 and SAM3. On COCO and LVIS, PicoSAM3 achieves 65.45% and 64.01% mIoU, respectively, outperforming existing SAM-based and edge-oriented baselines at similar or lower complexity. The INT8 quantized model preserves accuracy with negligible degradation while enabling real-time in-sensor inference at 11.82 ms latency on the IMX500, fully complying with its memory and operator constraints. Ablation studies show that distillation from large SAM models yields up to +14.5% mIoU improvement over supervised training and demonstrate that high-quality, spatially flexible promptable segmentation is feasible directly at the sensor level.
format Preprint
id arxiv_https___arxiv_org_abs_2603_11917
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PicoSAM3: Real-Time In-Sensor Region-of-Interest Segmentation
Bonazzi, Pietro
Farronato, Nicola
Zihlmann, Stefan
Qin, Haotong
Magno, Michele
Computer Vision and Pattern Recognition
Real-time, on-device segmentation is critical for latency-sensitive and privacy-aware applications such as smart glasses and Internet-of-Things devices. We introduce PicoSAM3, a lightweight promptable visual segmentation model optimized for edge and in-sensor execution, including deployment on the Sony IMX500 vision sensor. PicoSAM3 has 1.3 M parameters and combines a dense CNN architecture with region of interest prompt encoding, Efficient Channel Attention, and knowledge distillation from SAM2 and SAM3. On COCO and LVIS, PicoSAM3 achieves 65.45% and 64.01% mIoU, respectively, outperforming existing SAM-based and edge-oriented baselines at similar or lower complexity. The INT8 quantized model preserves accuracy with negligible degradation while enabling real-time in-sensor inference at 11.82 ms latency on the IMX500, fully complying with its memory and operator constraints. Ablation studies show that distillation from large SAM models yields up to +14.5% mIoU improvement over supervised training and demonstrate that high-quality, spatially flexible promptable segmentation is feasible directly at the sensor level.
title PicoSAM3: Real-Time In-Sensor Region-of-Interest Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.11917