Abstract:Visual sensors are widely deployed in many promising detection fields, such as precision measurement, visual object detection and search. Instance search is a special object search method based on visual sensor, which is of fundamental importance to many applications, including the area of civil and military. Although a quantity of instance search solutions has been proposed, it remains a critical challenge to provide a real-time, lightweight labeling, easy to deploy and able to simultaneous-object-positioning-and-classification method. To address the challenge, a novel instance search method named Faster-FIIS-GMS is described in this paper. The improved Faster R-CNN with the MRFR (Multi-Receptive Field Residual module) and the APM (Adaptive Proposals Mechanism) is proposed to achieve more accurate and time-saving object detection and background information elimination simultaneously. Next, a novel image matching method named Fast Intensity Initial Screening Algorithm (FIIS) is introduced to filter foreground information in an image preliminarily and then combined with Grid-based Motion Statistics (GMS) to achieve the final searching. The performance of Faster-FIIS-GMS is analyzed and validated by experimental tests in a dataset annotated by manual, and the experimental results demonstrate that the proposed method has great potential in instance search for real scene environment, as it achieves high accuracy and satisfactory real time performance using the general computer. We also provide the comparison to state-of-the-arts to investigate the superiority of the proposed scheme at the end of the work.

Faster-FIIS-GMS: a novel object detection framework for instance search

Foreground Gating and Background Refining Network for Surveillance Object Detection

IC-FPS: Instance-Centroid Faster Point Sampling Module for 3D Point-base Object Detection

Beyond Local Search: Tracking Objects Everywhere with Instance-Specific Proposals

Fast Object Retrieval Using Direct Spatial Matching

SGMFNet: a remote sensing image object detection network based on spatial global attention and multi-scale feature fusion

A Novel Wide-Area Multiobject Detection System with High-Probability Region Searching

Instance Mining with Class Feature Banks for Weakly Supervised Object Detection.

Multiple receptive fields and small-object-focusing weakly-supervised segmentation network for fast object detection

Lightweight Spatial Sliced-Concatenate-Multireceptive-Field Enhance and Joint Channel Attention Mechanism for Infrared Object Detection

Concurrent Detection and Recognition of Individual Object Based on Colour and P-Sift Features.

Small Object Detection Method Based on Adaptive Spatial Parallel Convolution and Fast Multi-Scale Fusion

Towards Accurate Localization by Instance Search

An effective method for small objects detection based on MDFFAM and LKSPP

An Interactively Reinforced Paradigm for Joint Infrared-Visible Image Fusion and Saliency Object Detection

Feature-fused SSD: fast detection for small objects

An Improved Faster R-CNN for Small Object Detection

Lightweight Real-Time Object Detection via Enhanced Global Perception and Intra-Layer Interaction for Complex Traffic Scenarios

Dynamic Feature and Context Enhancement Network for Faster Detection of Small Objects

Dataset-driven Unsupervised Object Discovery for Region-based Instance Image Retrieval

MSF: Motion-guided Sequential Fusion for Efficient 3D Object Detection from Point Cloud Sequences