Finding Four-Leaf Clovers: A Benchmark for Fine-Grained Object Localization

* Indicates equal contribution
We present the Four-Leaf Clover (FLC) dataset, a new experimental framework for studying fine-grained object localization problems. We built the FLC dataset with the contribution of trained hobbyists, who were assigned the task of spotting four-leaf clovers on a fixed geographical extension over two clover seasons, one season for the train set and another for the test set. We then annotated each object instance for the tasks of object detection, semantic segmentation, instance segmentation, object parsing and semantic boundary detection. Our dataset is composed of more than 100,000 images, containing 2,151 carefully annotated clover instances of four, five or six leaves. The FLC dataset is extremely challenging and adapted to fine-grained object localization problems due to its small inter-class variance and its very large intra-class variation. We perform extensive experiments with state-of-the-art methods in order to establish strong baselines for each of the tasks.
Figure 1. (Left panel) Example of an image of the FLC dataset. (Right panel) Positive samples of four-leaf clovers on theimage on the left.

Publications

Finding Four-Leaf Clovers: A Benchmark for Fine-Grained Object Localization

L Bravo*, A Pardo*, G Perez*, P Arbelaez

The Sixth Workshop on Fine-Grained Visual Categorization (FGVC6), CVPR 2019

The Four-Leaf Clover Dataset

Table 1. Comparison of FLC to major visual recognition datasets. Clubsuit (♣) indicates that a dataset allows to study a recognition problem at a fine-grained level, triangle (4) indicates that the version of the problem is not fine-grained, and (×) indicates that a dataset does not allow to study a problem. The first six rows correspond to object recognition datasets that lack a fine-grained nature, while the next five rows present examples of fine-grained image classification datasets.
Figure 2. Examples of segmentation annotations
Table 2. FLC dataset statistics. 4-leaf clover pixels and 4-leaf clover boundary pixels refer to the rate of the total of positive pixels over the total of pixels in the FLC dataset.

Tasks

Downloads

Citation

@article{BPPA_FGVC6,
author = "{L. Bravo, A. Pardo, G. Perez and P. Arbelaez}",
title = "{Finding Four-Leaf Clovers: A Benchmark for Fine-Grained Object Localization}",
journal = "{The Sixth Workshop on Fine-Grained Visual Categorization (FGVC6), CVPR 2019.}",
year = {2019},
}
  • Address:
    Cra. 1 E No. 19A - 40. 111711, Bogotá, Colombia - Mario Laserna Building - School of Engineering - Universidad de Los Andes
  • Phone:
    [571] 332 4327, 332 4328, 332 4329