A key task in computer vision called pedestrian detection includes finding and locating pedestrians in an image or video stream. Due to its applicability in many areas, such as driverless vehicles, surveillance, crowd control, and pedestrian safety systems, this problem has taken on a substantial amount of significance. The development of algorithms and models for independent pedestrian recognition that can form bounding boxes around pedestrians to precisely estimate their positions in visual input is the aim of pedestrian detection. There are various algorithms that are used in implementation of pedestrian detection but they suffer with various drawbacks like feature degradation and false positives from noise. Fully Convolutional One-Stage (FCOS) is an advanced object identification framework that works well for problems involving pedestrian detection. FCOS is very useful for handling overlapping, crowded objects like pedestrians. A deeper network like ResNet-101 can potentially learn more intricate and diverse representations from the different modalities, allowing it to capture and leverage the unique characteristics present in each spectral band more effectively. Considering the benefits of FCOS and ResNet-101, The proposed model which is a combination of FCOS and resnet-101 to overcome the challenges in pedestrian detection. The proposed model performs better than state-of-art methods in terms of precision, recall and F1 Score.
Robust Pedestrian Detection in Challenging Environmental Conditions Using FCOS
2023-12-04
794404 byte
Conference paper
Electronic Resource
English
NFIS: A NMS-free FCOS Method for Instance Segmentation
IEEE | 2021
|Vision-Based Robust Lane Detection and Tracking under Different Challenging Environmental Conditions
ArXiv | 2022
|Robust Vehicle Detection and Distance Estimation Under Challenging Lighting Conditions
Online Contents | 2015
|