Can we trust bounding box annotations for object detection?

Murrugarra-Llerena, Jeffri; Kirsten, Ln; Jung, Cláudio Rosito

doi:10.1109/cvprw56347.2022.00528

Cited by 5 publications

(2 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…8). As such, the ground-truth annotations are expected to comprise a "human variability" factor [25], which can also impact the objective metrics used to compare object detection approaches [27]. In this section, we propose to evaluate the level of agreement among human annotators regarding cell detection and classification (a.k.a.…”

Section: B Inter-annotator Assessmentmentioning

confidence: 99%

“…We also provide a smaller subset of images annotated by five different human experts and perform a quantitative analysis of inter-human annotation agreement. As noted in [25], even small discrepancies in bounding box annotations can lead to strong IoU degradation. In this work, we show that this behavior is amplified in cell microscopy imagery due to the inherent difficulty of the problem.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Oriented Cell Dataset: efficient imagery analyses using angular representation

Kirsten,

Angonezi,

Oliveira

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

In this work, we propose a new public dataset for cell detection in bright-field microscopy images annotated with Oriented Bounding Boxes (OBBs), named Oriented Cell Dataset (OCD). We show that OBBs provide a more accurate shape representation compared to standard Horizontal Bounding Boxes (HBBs), with slight overhead of one extra click in the annotation process. Our dataset also contains a subset of images with five independent expert annotations, which allows inter-annotation analysis to determine if the results produced by algorithms are within the expected variability of human experts. We investigated how to automate cell biology microscopy images by training seven popular OBB detectors in the proposed dataset, and focused our analyses on two main problems in cancer biology: cell confluence and polarity determination, the latter not possible through HBB representation. All models achieved statistically similar results to the biological applications compared to human annotation, enabling the automation of cell biology and cancer cell biology microscopy image analysis. Our code and dataset are available at https://github.com/LucasKirsten/Deep-Cell-Tracking-EBB.

show abstract

Section: B Inter-annotator Assessmentmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Oriented Cell Dataset: efficient imagery analyses using angular representation

Kirsten,

Angonezi,

Oliveira

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Automated Detection of Hillforts in Remote Sensing Imagery With Deep Multimodal Segmentation

Canedo,

Fonte,

Dias

et al. 2024

Archaeological Prospection

View full text Add to dashboard Cite

Recent advancements in remote sensing and artificial intelligence can potentially revolutionize the automated detection of archaeological sites. However, the challenging task of interpreting remote sensing imagery combined with the intricate shapes of archaeological sites can hinder the performance of computer vision systems. This work presents a computer vision system trained for efficient hillfort detection in remote sensing imagery. Equipped with an adapted multimodal semantic segmentation model, the system integrates LiDAR‐derived LRM images and aerial orthoimages for feature fusion, generating a binary mask pinpointing detected hillforts. Post‐processing includes margin and area filters to remove edge inferences and smaller anomalies. The resulting inferences are subjected to hard positive and negative mining, where expert archaeologists classify them to populate the training data with new samples for retraining the segmentation model. As the computer vision system is far more likely to encounter background images during its search, the training data are intentionally biased towards negative examples. This approach aims to reduce the number of false positives, typically seen when applying machine learning solutions to remote sensing imagery. Northwest Iberia experiments witnessed a drastic reduction in false positives, from 5678 to 40 after a single hard positive and negative mining iteration, yielding a 99.3% reduction, with a resulting F1 score of 66%. In England experiments, the system achieved a 59% F1 score when fine‐tuned and deployed countrywide. Its scalability to diverse archaeological sites is demonstrated by successfully detecting hillforts and other types of enclosures despite their typical complex and varied shapes. Future work will explore archaeological predictive modelling to identify regions with higher archaeological potential to focus the search, addressing processing time challenges.

show abstract