The Challenge of Data Annotation in Deep Learning—A Case Study on Whole Plant Corn Silage

Abstract

Recent advances in computer vision are primarily driven by the usage of deep learning, which is known to require large amounts of data, and creating datasets for this purpose is not a trivial task. Larger benchmark datasets often have detailed processes with multiple stages and users with different roles during annotation. However, this can be difficult to implement in smaller projects where resources can be limited. Therefore, in this work we present our processes for creating an image dataset for kernel fragmentation and stover overlengths in Whole Plant Corn Silage. This includes the guidelines for annotating object instances in respective classes and statistics of gathered annotations. Given the challenging image conditions, where objects are present in large amounts of occlusion and clutter, the datasets appear appropriate for training models. However, we experience annotator inconsistency, which can hamper evaluation. Based on this we argue the importance of having an evaluation form independent of the manual annotation where we evaluate our models with physically based sieving metrics. Additionally, instead of the traditional time-consuming manual annotation approach, we evaluate Semi-Supervised Learning as an alternative, showing competitive results while requiring fewer annotations. Specifically, given a relatively large supervised set of around 1400 images we can improve the Average Precision by a number of percentage points. Additionally, we show a significantly large improvement when using an extremely small set of just over 100 images, with over 3× in Average Precision and up to 20 percentage points when estimating the quality.

Digital Phenotyping in Plant Breeding: Evaluating Relative Maturity, Stand Count, and Plant Height in Dry Beans (Phaseolus vulgaris L.) via RGB Drone-Based Imagery and Deep Learning Approaches

“…4 -trend lines). Compared to the image processing methods used in this study (OpenCV and WS), the annotation quality is an important issue when performing object detection analysis [105]. This observation underscores the importance of accurate data collection and annotation to ensure reliable and robust results in object detection studies.…”

Combining DEVS simulation and ontological modeling for hierarchical analysis of the SARS-CoV-2 replication

“…This leads scientists to train their models on unlabeled data sets, instead of annotated and important data. 66–68 Furthermore, ML approaches require much data, however, there are not enough. 69…”

“…This leads scientists to train their models on unlabeled data sets, instead of annotated and important data. [66][67][68] Furthermore, ML approaches require much data, however, there are not enough. 69 Besides, the majority of these studies focus on the relative risk of SARS-CoV-2 and its transmission such as investigating molecular dynamics simulations, reproducing the viral infection, measuring the effects of drug treatment, identifying potential drug candidates, spreading of the virus through people, and so forth.…”

Detection of abnormal extraocular muscles in small datasets of computed tomography images using a three-dimensional variational autoencoder

“…First, it is difficult to collect sufficient medical image datasets. Second, few annotated medical datasets are available because they require a laborious annotation process to be performed by the trained experts and are restricted by legal issues associated with public access to private medical information 6 , 7 . Because supervised DL generally relies on large amounts of data that have been accurately annotated by experts, the data preparation phase of medical imaging studies can be challenging.…”