Skip to main navigation Skip to search Skip to main content

Robust Multi-Scale Object Segmentation in Surveillance Video

    Activity: Supervision › Completed SURF Project

    Description

    Surveillance imagery poses significant challenges for object detection and
    segmentation, particularly for small and difficult targets.This work investigates deep-learning-based object detection and instancesegmentation methods, comparing model performance across differentbiect scales.Evaluation is conducted using loU and mAP, with emphasis on small and hardtargets. The goal is to develop robust deep-learning approaches for multi-scale object understanding in surveillance scenarios.
    Period29 Jun 2026 → 29 Aug 2026

    Keywords

    • Deep Learning
    • Instance Segmentatio
    • Small Object Detectio
    • Multi-scale Object Understandin
    • Surveillance Scenari