ORV: A comprehensive dataset of on-road vehicle images

IntroductionORV (On-Road Vehicle) dataset is built under the needs of image analysis of on-road vehicles, aiming at extracting the detailed information of vehicles in an image. The general information of a vehicle contains the location in the image, color, type, brand, plate number, etc. In order to address such kind of variousness, we construct five different datasets: one mainly for the localization task and the other four are for the recognition tasks. Specifically, the dataset for localization is annotated with the bounding boxes and the type of each vehicle. The instances in the rest datasets are classified into different folder to represent their categories. For example, the vehicle brand dataset contains 79 folders and each folder corresponds to a vehicle brand. The following chapter will show you the architecture of these datasets and some examples of each dataset. Users could refer to the tutorial of each dataset for further details. Structure of Five SubsetsLocalization/DetectionFigure 1: Structure of the detection dataset and exemplar annotated image.BrandFigure 2: Structure of the vehicle brand dataset and exemplar images.ColorFigure 3: Structure of the vehicle color dataset and exemplar images.Plate ClassificationFigure 4: Structure of the license plate dataset and exemplar images.License Plate Recognition (LPR)Figure 5: Structure of the license plate recognition (LPR) dataset and exemplar images.DownloadYou can download the datasets using the following links: (please contact VIM research group for more information.)Original Images and AnnotationVehicle BrandVehicle ColorLicense Plate JudgementLicense Plate Recognition (LPR)CitationIf you use this dataset in your publication, please cite the dataset and following paper:

Read More

VegFru: A Domain-Specific Dataset for Fine-grained Visual Categorization

AbstractIn this paper, we propose a novel domain-specific dataset named VegFru for fine-grained visual categorization (FGVC). While the existing datasets for FGVC are mainly focused on animal breeds or man-made objects with limited labelled data, VegFru is a larger dataset consisting of vegetables and fruits which are closely associated with the daily life of everyone. Aiming at domestic cooking and food management, VegFru categorizes vegetables and fruits according to their eating characteristics, and each image contains at least one edible part of vegetables or fruits with the same cooking usage. Particularly, all the images are labelled hierarchically. The current version covers vegetables and fruits of 25 upper-level categories and 292 subordinate classes. And it contains more than 160,000 images in total and at least 200 images for each subordinate class. Accompanying the dataset, we also propose an effective framework called HybridNet to exploit the label hierarchy for FGVC. Specifically, multiple granularity features are first extracted by dealing with the hierarchical labels separately. And then they are fused through explicit operation, \eg, Compact Bilinear Pooling, to form a unified representation for the ultimate recognition. The experimental results on the novel VegFru, the public FGVC-Aircraft and CUB-200-2011 indicate that HybridNet achieves one of the top performance on these datasets.1. VegFru1.1 OverviewFigure 1: Sample images in VegFru. Top: vegetable images. Bottom: fruit images. Best viewed electronically.Currently, the dataset covers vegetables and fruits of 25 upper-level categories (denoted as sup-class) and 292 subordinate classes (denoted as sub-class), which has taken in all species in common. It contains more than 160,000 images in total and at least 200 images for each sub-class, which is much larger than the previous fine-grained datasets. Particularly, besides the fine-grained annotation, the images in VegFru are assigned hierarch

Read More