Abstract:Object detection via deep neural networks has undergone considerable advancements in recent years. Yet, the detection of smaller objects, specifically those with a few pixels (i.e., < 322 pixels), is still challenging compared with large objects (i.e., > 962 pixels). Existing methods commonly apply high-resolution features or complex super-resolution strategies based on the two-stage Faster Region Convolutional Neural Network (RCNN). They sequentially apply localization and classification stages after a shared feature map extracted by one single backbone network. However, these methods cause low detection accuracy of small objects, high computational overhead, and waste of hardware resources. In this paper, we develop a high-accuracy and real-time small object detection system with negligible computational overhead and low hardware idleness. At the software level, we propose a two-stage Coarse-to-Fine Decoupling RCNN (CFD RCNN) with three techniques: (1) The shared backbone decoupling for localization and classification to achieve high accuracy for both tasks; (2) The training method using backbone feature upsampling for localization with low computational overhead; (3) The object cropping strategy from the original high-resolution image for high-accuracy classification. At the hardware level, we propose a virtualized FPGA accelerator with the Dynamic Resource Allocation (DRA) strategy. The DRA strategy reallocates the hardware resources, considering the workload and resource preference of each stage in CFD RCNN to reduce hardware idleness. Extensive experiments on the TT100K and GTSDB datasets using Xilinx ZCU102 FPGA show that the proposed small object detection system can achieve 2.9% improvement in mean average precision (mAP) compared with state-of-the-art (SOTA) algorithms and raised the throughput from 18.9 FPS to > 26.0 FPS (~1.37×) compared with existing accelerators.

Binarization-Based Human Detection with Hardware Reconfigurability

Adaptive Human Detection Approach Using FPGA-based Parallel Architecture in Reconfigurable Hardware.

High-performance Reconfigurable DNN Accelerator on a Bandwidth-limited Embedded System

An Intelligent Surveillance System Base on Accelerated Human and Face Detection

FPGA Implementation of Feature Detection Algorithm Based on High Level Synthesis

Hardware–Software Partitioning for Real-Time Object Detection Using Dynamic Parameter Optimization

Design of embedded real-time human detection system based on Zynq platform

Efficient FPGA Implementation of Human Detection from Video Sequences

Fast AdaBoost-Based Face Detection System on a Dynamically Coarse Grain Reconfigurable Architecture.

Hardware Architecture for Fast General Object Detection using Aggregated Channel Features

A Real-Time Human Detection System for Video.

Towards High-accuracy and Real-time Two-stage Small Object Detection on FPGA

An Fpga-Based Dynamic Reconfigurable Cnn Edge Computing Device for Object Classification

Urinary d-glucaric acid excretion and serum gamma-glutamyl transpeptidase activity in alcoholism.

Edge Real-Time Object Detection and DPU-Based Hardware Implementation for Optical Remote Sensing Images

Hardware Implementation of Depthwise Separable Convolution Neural Network

Real-time low-power binocular stereo vision based on FPGA

An Efficient Real-Time Object Detection Framework on Resource-Constricted Hardware Devices via Software and Hardware Co-design

Reduced-Parameter YOLO-Like Object Detector Oriented to Resource-Constrained Platform

A High Speed Reconfigurable Face Detection Architecture

Computational models of object motion detectors accelerated using FPGA technology