Plum Detection For Harvesting
This site contains the plum detection dataset from the plum harvesting prototype trial described in 'Design and Evaluation of a Modular Robotic Plum Harvesting System Utilising Soft Components' and the raw sensor recordings can be found here.
The dataset itself is here and can be downloaded using bash with:
'wget --recursive --no-parent -R "index.html*" http://data.acfr.usyd.edu.au/Agriculture/PlumDetection/dataset/'
The benchmark networks trained on this data and their training parameters are available here and should be used with the github repos:
https://github.com/jaspereb/tf-faster-rcnn
https://github.com/jaspereb/keras-yolo3
https://github.com/jaspereb/keras-retinanet
https://github.com/jaspereb/CenterNet
The dataset was generated by the agriculture team at the Australian Centre for Field Robotics, The University of Sydney, Australia. Further information about the group can be found at sydney.edu.au/acfr/agriculture
Data summary
This dataset contains a day and night version, in the Pascal VOC format. There are 350 images in each set, with a total of 4449 and 1402 red plum instances in the day and night set repsectively. Default splits in the ImageSets > Main folder.
The VOC 'difficult' tag is False for all annotations, while 'truncated' is accurate (True if the bounding box hits the edge of the image). The 'DepthImages' folder contains depth maps corresponding to each JPEG image in the .npy format. These have been processed as described in the paper, by clipping depth to between 0.11m and 2.5m. They are then normalised by dividing the depth by 2.5m.
The 'DepthMats' folder contains the same data, but without the clipping and normalisation. This is in the .mat format so can be easily inspected using matlab or scipy.
Examples Day Images; Top row is RGB image, Bottom row is unclipped depth image, Middle row is an overlay of the two.
Examples Night Images; Top row is RGB image, Bottom row is unclipped depth image, Middle row is an overlay of the two.
For any questions, feel free to contact j.brown@acfr.usyd.edu.au
