Abstract
Training deep learning models typically requires a huge amount of labeled data which is expensive to acquire, especially in dense prediction tasks such as semantic segmentation. Moreover, plant phenotyping datasets pose additional challenges of heavy occlusion and varied lighting conditions which makes annotations more time-consuming to obtain. Active learning helps in reducing the annotation cost by selecting samples for labeling which are most informative to the model, thus improving model performance with fewer annotations. Active learning for semantic segmentation has been well studied on datasets such as PASCAL VOC and Cityscapes. However, its effectiveness on plant datasets has not received much importance. To bridge this gap, we empirically study and benchmark the effectiveness of four uncertainty-based active learning strategies on three natural plant organ segmentation datasets. We also study their behaviour in response to variations in training configurations in terms of augmentations used, the scale of training images, active learning batch sizes, and train-validation set splits.
Highlights
Deep learning models have been widely used for various plant phenotyping tasks by formulating them as standard vision tasks such as image classification [1,2,3,4], object detection [5,6,7,8], and semantic segmentation [9,10,11]
We believe this is the case due to the similarity between informativeness scores calculated by the query metrics at hand, largely contributed by the binary nature of the task, which is clearly explained in Chapter 3 of [17]
We have studied the efficiency of uncertaintybased Active learning (AL) strategies extensively on three plant organ segmentation datasets
Summary
Deep learning models have been widely used for various plant phenotyping tasks by formulating them as standard vision tasks such as image classification [1,2,3,4], object detection [5,6,7,8], and semantic segmentation [9,10,11]. Owing to the success of data-driven learning methods, many practitioners have relied on collecting, labeling, and maintaining large amounts of data to solve tasks of their interest This popularized pool-based variant of AL methods which allow for selecting active samples from a large pool of unlabeled data. Image resolution plays a crucial role while impact of changes in experimental conditions such as data training deep neural networks.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have