An Image-Based Benchmark Dataset and a Novel Object Detector for Water Surface Object Detection

Abstract

Water surface object detection is one of the most significant tasks in autonomous driving and water surface vision applications. To date, existing public large-scale datasets collected from websites do not focus on specific scenarios. As a characteristic of these datasets, the quantity of the images and instances is also still at a low level. To accelerate the development of water surface autonomous driving, this paper proposes a large-scale, high-quality annotated benchmark dataset, named Water Surface Object Detection Dataset (WSODD), to benchmark different water surface object detection algorithms. The proposed dataset consists of 7,467 water surface images in different water environments, climate conditions, and shooting times. In addition, the dataset comprises a total of 14 common object categories and 21,911 instances. Simultaneously, more specific scenarios are focused on in WSODD. In order to find a straightforward architecture to provide good performance on WSODD, a new object detector, named CRB-Net, is proposed to serve as a baseline. In experiments, CRB-Net was compared with 16 state-of-the-art object detection methods and outperformed all of them in terms of detection precision. In this paper, we further discuss the effect of the dataset diversity (e.g., instance size, lighting conditions), training set size, and dataset details (e.g., method of categorization). Cross-dataset validation shows that WSODD significantly outperforms other relevant datasets and that the adaptability of CRB-Net is excellent.

A Guide to Image and Video based Small Object Detection using Deep Learning : Case Study of Maritime Surveillance

“…Tian et al [166] embeded multiple Atrous Spatial Pyramid Pooling (ASPP) modules in FPN to improve the detection performance for ships at different scales. Zhou et al [182] proposed CRB-Net, a multi-scale image feature learning based method that can carry out adaptive weight adjustment (improved BIFPN) during feature fusion by attention mechanism and Mish activation (a novel self-regularized non-monotonic activation function [183]). Two SPPNets were also used to increase the receptive field of the features in layers 4 and 5 to isolate the most significant contextual features.…”

“…WSODD Dataset [182] or Water Surface Object Detection dataset was developed for obstacle detection on water surfaces. The images include oceans, rivers, and lakes that were acquired at different times and weather conditions, such as during the day, twilight, and night, sunny, cloudy, or foggy conditions.…”

Self-driving Vessels: YOLOv5 Approach for Water Surface Object Detection

“…Ainda que haja promissores detectores de objetos em estado da arte disponíveis conforme mencionado acima, há uma significativa escassez de abordagens baseadas em arquiteturas convolucionais que são especializadas na detecc ¸ão de objetos localizados em superfície de água [Prasad et al 2017]. Isso ocorre principalmente em virtude de existirem poucas bases de imagens disponíveis publicamente que retratam objetos em superfície de água [Zhou et al 2021]. Das poucas bases disponíveis, a maioria é caracterizada por não ser representativa em relac ¸ão à quantidade de instâncias por categoria.…”

“…Recentemente, foi disponibilizada uma base de imagens anotadas que retratam, com notável qualidade e variedade, objetos localizados em superfície de água. Essa base, intitulada como WSODD (Water Surface Object Detection Dataset) [Zhou et al 2021], mitiga os pontos levantados anteriormente e será utilizada como referência para comparar a performance do detector YOLOv5 com diferentes detectores clássicos de objetos. Considerando o que foi exposto acima, as principais contribuic ¸ões deste artigo são:…”

“…(II) Comparar a performance do detector YOLOv5 com outros diferentes métodos de detecc ¸ão apresentados no trabalho de [Zhou et al 2021], utilizando a mesma base de imagens WSODD. Ao final, observou-se que o método YOLOv5 superou significativamente todas as outras abordagens comparadas.…”