capcelera

 

M-UdL_200

 

 

Logo Agrotecnio

 

 

 Centres

ETSEA


 

 

Departaments

EAGROF

 

 

 

 

 

 

Inici > English version > GRAP publications > PApple_RGB-D-Size dataset

PApple_RGB-D-Size dataset


The PApple_RGB-D-Size dataset is composed by 4017 RGB-D images of 6 Fuji apples trees with ground truth annotations for training and testing apple segmentation and diameter estimation algorithms. Three of the imaged trees were at an advanced ripening stage (reddish apples), while the other three trees were scanned when the apples were at the 70% of their final size (green apples).

Color images are saved in .png format (inside “\images” folder), while depth images are saved in .npy format (inside “\depthCropNpy” folder). These depth images were generated using structure-from-motion and multi-view stereo techniques.

All images were annotated with apple segmentation masks and apples diameter ground truth. Instance segmentation masks are saved in .json format (inside “\gt_json” folder). In addition, the “GT_diameter.txt” file (inside “gt_json” folder) provides the correspondence between each apple ID and the ground truth apple diameter. To visualize instance segmentation masks and the corresponding apple IDs authors suggest to use the VIA annotator tool.

PApple_RGB-D-Size_dataset

Figure 1: Sample of 3 RGB-D images extracted from test dataset and their associated fruit segmentation and diameter ground truth annotations. First column corresponds to colour images (RGB), second column to Depth images and the third column to the RGB annotated images.

 

The database can be downloaded from the following link: PApple_RGB-D-Size dataset (11.5 GB)

Find the baseline used for fruit detection in [1] at https://github.com/GRAP-UdL-AT/multitask_RGB-D_FruitDetectionAndSizing

 

If you use the database in any publications or reports, please, refer to the following papers:

[1] Ferrer Ferrer M, Ruiz-Hidalgo J, Gregorio E, Vilaplana V, Morros JR, Gené-Mola J. 2022. Simultaneous Fruit Detection and Size Estimation Using Multitask Deep Neural Networks. (Submitted)

 

 

Last modified: 11/03/2022
Imprimir Enrera Pujar


Home | Accessibility | About the web | Site map | Legal notice | Contact
© 2014 Universitat de Lleida - Pl. de Víctor Siurana 1, 25003 Lleida - Tel. (+34) 973 702 000 - All rights reserved
Ombra