Paper V4A2024

DISCLAIMER

This dataset is provided "as is" without any warranty express or implied. The provider disclaims all warranties including but not limited to warranties of merchantability, fitness for a particular purpose, and non-infringement.


ATTRIBUTION

If you use this dataset in your research, please cite "Gonçalo P. Matos, Carlos Santiago, João Costeira, Ricardo L. Saldanha and Ernesto M. Morgado, Tracking and Counting Apples in Orchards Under Intermittent Occlusions and Low Frame Rates, CVPR 2024 Vision for Agriculture Workshop, Seattle, 2024".
You should also mention this dataset and its source in any publications resulting from your work.


DATASET DESCRIPTION

The data published here corresponds to four datasets for apple detection and tracking / counting.

The first three datasets correspond to video sequences obtained with a manually operated camera, moved by a person on foot, in an apple orchard from INIAV in Alcobaça, Portugal. The captured trees correspond to three different varieties of apples - named Galafab, Schnico Red and Schniga Schnico - in the final stages of maturity, close to harvest.
The fourth dataset was synthetically generated at SISCOG, and corresponds to 5 trees containing apples, "filmed" with a stabilized, horizontally moving camera.

For the first three datasets, the video frames are provided in two formats:

  • The original frames, as captured by the camera. These were used in the paper to perform the 3D reconstructions with SfM software;
  • The masked frames, where some trees where painted in black. These were used in the paper to count fruits, to match the trees that were manually counted in the field. If you want to replicate / compare with the counting results published in the paper, use these frames.

For the synthetic dataset, ground truth annotations with the fruits' bounding boxes and track IDs are provided. This allows for testing other tracking and counting algorithms.

For more information about the datasets, refer to the paper and its corresponding supplemental material published at CVPR.