Dataset Ninja LogoDataset Ninja:

BUP19 Dataset

57241050
Tagagriculture, robotics
Taskinstance segmentation
Release YearMade in 2019
Licenseunknown

Introduction #

Michael Halstead, Simon Denman, Clinton Fookeset al.

The authors release BUP19: Sweet Pepper Dataset to explore the issue of generalisability of the large differences between cropping environments by considering a fruit (sweet pepper) that is grown using different cultivars (sub-species) and in different environments (field vs glasshouse). The authors perform an evaluation of sweet pepper detection, segmentation, and quality assessment. Evaluations show that while domain specific models yield higher performance on their source data, detection and segmentation of sweet pepper on unseen data is viable for agricultural robotics by exploiting multi-task learning.

Note, similar BUP19: Sweet Pepper Dataset datasets are also available on the DatasetNinja.com:

Motivation

The field of agricultural robotics is experiencing rapid growth, driven by advancements in computer vision, machine learning, and robotics, coupled with an increasing demand in agriculture. Despite these strides, a significant gap exists between the technology available and the diverse requirements of farming, primarily due to the substantial variations in cropping environments. This underscores the urgent need for models with enhanced generalizability.

Addressing this gap, the BUP19 dataset was meticulously curated, encompassing diverse domains, cultivars, cameras, and geographic locations. The authors leverage this dataset in both individual and combined approaches to assess the detection and classification of sweet peppers in the wild. They opt for Faster-RCNN for detection, benefiting from its seamless adaptability to multitask learning through the incorporation of the Mask-RCNN framework for instance-based segmentation.

In evaluating sub-class classification, with a focus on the accuracy of correct detections, the authors achieve an impressive accuracy score of 0.9 in cross-domain evaluations. Notably, their experiments reveal that intra-environmental inference tends to be suboptimal. However, by enriching the data through a combination of diverse datasets, the authors enhance performance by introducing greater diversity into the training data.

In summary, the presentation of unique and varied datasets exemplifies the capacity of multi-task learning to improve cross-dataset generalization. Concurrently, it emphasizes the crucial role of diverse data in the efficient training and evaluation of real-world systems.

Dataset description

The BUP19 dataset was captured in a glass house replicating a commercial setting at Campus Klein-Altendorf. Two different cultivar of sweet pepper were grown simultaneously during experiments: Mazurka (Rijk Zwaan) and Mavras (Enza Zaden). The glass house for sweet pepper cultivation was arranged into six rows of approximately 40m in length each. Data was recorded into bagfiles using an Intel RealSense D435i camera at 30fps. For recording each row was separated into
four equally spaced sections. Post processing was completed to align the depth and RGB images using the pyrealsense23 libraries. The stored depth image is a uint16 TIFF format file where 1mm is represented by each change in value.

image

Example images from the BUP19 dataset: (a) is the raw image, (b) is a colourised version of the instance masks, (c)-(f) are representations of the instance masks for black, green, mixed, and red, and (g) is a quantized version of the depth image for visualisation.

For annotation, the glasshouse data was separated into three distinct sections: 1/3 training, 1/3 validation, and 1/3 evaluation. The separation of sections during recordings allowed for the data to be evenly split between each sub-set. Extending beyond bounding box regression, instance based masks are annotated. Annotation was completed by three individuals who annotated different images. A separate mask is included for each sub-class where zero denotes “background”, and a numbered response indicates the presence of a sweet pepper.

ExpandExpand
Dataset LinkHomepageDataset LinkResearch Paper

Summary #

BUP19: Sweet Pepper Dataset is a dataset for instance segmentation, semantic segmentation, object detection, and monocular depth estimation tasks. It is used in the agricultural and robotics industries.

The dataset consists of 572 images with 7448 labeled objects belonging to 4 different classes including black pepper, green pepper, mixed pepper, and other: red pepper.

Images in the BUP19 dataset have pixel-level instance segmentation and bounding box annotations. Due to the nature of the instance segmentation task, it can be automatically transformed into a semantic segmentation task (only one mask for every class). All images are labeled (i.e. with annotations). There are 3 splits in the dataset: train (228 images), test (176 images), and val (168 images). Additionally, images are grouped by im id. Also, every image in train dataset marked with alireza or claus tag. The dataset was released in 2019 by the University of Bonn, Germany.

Dataset Poster

Explore #

BUP19 dataset has 572 images. Click on one of the examples below or open "Explore" tool anytime you need to view dataset images with annotations. This tool has extended visualization capabilities like zoom, translation, objects table, custom filters and more. Hover the mouse over the images to hide or show annotations.

OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
OpenSample annotation mask from BUP19Sample image from BUP19
👀
Have a look at 572 images
Because of dataset's license preview is limited to 12 images
View images along with annotations and tags, search and filter by various parameters

Class balance #

There are 4 annotation classes in the dataset. Find the general statistics and balances for every class in the table below. Click any row to preview images that have labels of the selected class. Sort by column to find the most rare or prevalent classes.

Search
Rows 1-4 of 4
Class
Images
Objects
Count on image
average
Area on image
average
black pepper
any
510
4160
8.16
2.07%
green pepper
any
458
2316
5.06
1.37%
mixed pepper
any
336
612
1.82
0.86%
red pepper
any
238
360
1.51
0.55%

Co-occurrence matrix #

Co-occurrence matrix is an extremely valuable tool that shows you the images for every pair of classes: how many images have objects of both classes at the same time. If you click any cell, you will see those images. We added the tooltip with an explanation for every cell for your convenience, just hover the mouse over a cell to preview the description.

Images #

Explore every single image in the dataset with respect to the number of annotations of each class it has. Click a row to preview selected image. Sort by any column to find anomalies and edge cases. Use horizontal scroll if the table has many columns for a large number of classes in the dataset.

Object distribution #

Interactive heatmap chart for every class with object distribution shows how many images are in the dataset with a certain number of objects of a specific class. Users can click cell and see the list of all corresponding images.

Class sizes #

The table below gives various size properties of objects for every class. Click a row to see the image with annotations of the selected class. Sort columns to find classes with the smallest or largest objects or understand the size differences between classes.

Search
Rows 1-4 of 4
Class
Object count
Avg area
Max area
Min area
Min height
Min height
Max height
Max height
Avg height
Avg height
Min width
Min width
Max width
Max width
black pepper
any
4160
0.25%
2.05%
0%
6px
0.47%
701px
54.77%
60px
4.67%
5px
0.69%
720px
100%
green pepper
any
2316
0.27%
1.81%
0.01%
12px
0.94%
172px
13.44%
61px
4.78%
9px
1.25%
394px
54.72%
mixed pepper
any
612
0.47%
1.77%
0.01%
12px
0.94%
639px
49.92%
81px
6.35%
13px
1.81%
570px
79.17%
red pepper
any
360
0.36%
2.17%
0.01%
10px
0.78%
621px
48.52%
70px
5.49%
11px
1.53%
518px
71.94%

Spatial Heatmap #

The heatmaps below give the spatial distributions of all objects for every class. These visualizations provide insights into the most probable and rare object locations on the image. It helps analyze objects' placements in a dataset.

Spatial Heatmap

Objects #

Table contains all 7448 objects. Click a row to preview an image with annotations, and use search or pagination to navigate. Sort columns to find outliers in the dataset.

Search
Rows 1-10 of 7448
Object ID
Class
Image name
click row to open
Image size
height x width
Height
Height
Width
Width
Area
1
black pepper
any
frame_2019_10_1_8_39_0_205684.png
1280 x 720
92px
7.19%
32px
4.44%
0.23%
2
green pepper
any
frame_2019_10_1_8_39_0_205684.png
1280 x 720
115px
8.98%
93px
12.92%
1%
3
green pepper
any
frame_2019_10_1_8_39_0_205684.png
1280 x 720
98px
7.66%
84px
11.67%
0.5%
4
green pepper
any
frame_2019_10_1_8_39_0_205684.png
1280 x 720
104px
8.12%
87px
12.08%
0.65%
5
black pepper
any
frame_2019_10_29_8_48_29_015887.tiff
1280 x 720
44px
3.44%
15px
2.08%
0.05%
6
black pepper
any
frame_2019_10_29_8_48_29_015887.tiff
1280 x 720
87px
6.8%
83px
11.53%
0.51%
7
green pepper
any
frame_2019_10_29_8_48_29_015887.tiff
1280 x 720
84px
6.56%
63px
8.75%
0.39%
8
green pepper
any
frame_2019_10_29_8_48_29_015887.tiff
1280 x 720
74px
5.78%
64px
8.89%
0.3%
9
mixed pepper
any
frame_2019_10_29_8_48_29_015887.tiff
1280 x 720
111px
8.67%
83px
11.53%
0.77%
10
mixed pepper
any
frame_2019_10_29_8_48_29_015887.tiff
1280 x 720
31px
2.42%
57px
7.92%
0.07%

License #

License is unknown for the BUP19: Sweet Pepper Dataset dataset.

Source

Citation #

If you make use of the BUP19 data, please cite the following reference:

@inproceedings{halstead2020fruit,
  title={Fruit detection in the wild: The impact of varying conditions and cultivar},
  author={Halstead, Michael and Denman, Simon and Fookes, Clinton and McCool, Chris},
  booktitle={2020 Digital Image Computing: Techniques and Applications (DICTA)},
  pages={1--8},
  year={2020},
  organization={IEEE}
}

Source

If you are happy with Dataset Ninja and use provided visualizations and tools in your work, please cite us:

@misc{ visualization-tools-for-bup19-dataset,
  title = { Visualization Tools for BUP19 Dataset },
  type = { Computer Vision Tools },
  author = { Dataset Ninja },
  howpublished = { \url{ https://datasetninja.com/bup19 } },
  url = { https://datasetninja.com/bup19 },
  journal = { Dataset Ninja },
  publisher = { Dataset Ninja },
  year = { 2024 },
  month = { oct },
  note = { visited on 2024-10-15 },
}

Download #

Please visit dataset homepage to download the data.

. . .

Disclaimer #

Our gal from the legal dep told us we need to post this:

Dataset Ninja provides visualizations and statistics for some datasets that can be found online and can be downloaded by general audience. Dataset Ninja is not a dataset hosting platform and can only be used for informational purposes. The platform does not claim any rights for the original content, including images, videos, annotations and descriptions. Joint publishing is prohibited.

You take full responsibility when you use datasets presented at Dataset Ninja, as well as other information, including visualizations and statistics we provide. You are in charge of compliance with any dataset license and all other permissions. You are required to navigate datasets homepage and make sure that you can use it. In case of any questions, get in touch with us at hello@datasetninja.com.