Revisiting the CompCars Dataset for Hierarchical Car Classification: New Annotations, Experiments, and Results

Buzzelli, Marco; Segantin, Luca

doi:10.3390/s21020596

Cited by 12 publications

(6 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• SoftMax: unilized for multi-classification functions; sometimes called the normalized exponential function [10]. Normalizes the input into a probability distribution that sums to 1.…”

Section: Theoretical Background 1sigmoid and Softmax Activation Funct...mentioning

confidence: 99%

Adapted CNN-SMOTE-BGMM Deep Learning Framework for Network Intrusion Detection using Unbalanced Dataset

Kamil,

Mohammed

2023

Iraqi Journal of Science

View full text Add to dashboard Cite

This paper proposes a new methodology for improving network security by introducing an optimised hybrid intrusion detection system (IDS) framework solution as a middle layer between the end devices. It considers the difficulty of updating databases to uncover new threats that plague firewalls and detection systems, in addition to big data challenges. The proposed framework introduces a supervised network IDS based on a deep learning technique of convolutional neural networks (CNN) using the UNSW-NB15 dataset. It implements recursive feature elimination (RFE) with extreme gradient boosting (XGB) to reduce resource and time consumption. Additionally, it reduces bias towards the majority class of the dataset by combining the Synthetic Minority Oversampling Technique (SMOTE) with the Bayesian Gaussian Mixture Model (BGMM) to solve the data imbalance problem. The results demonstrate that this model greatly outperforms the existing approaches, attaining identification rates in the binary classification of up to 98.80% and the multiple group classification of up to 96.49%.

show abstract

“…• SoftMax: unilized for multi-classification functions; sometimes called the normalized exponential function [10]. Normalizes the input into a probability distribution that sums to 1.…”

Section: Theoretical Background 1sigmoid and Softmax Activation Funct...mentioning

confidence: 99%

Adapted CNN-SMOTE-BGMM Deep Learning Framework for Network Intrusion Detection using Unbalanced Dataset

Kamil,

Mohammed

2023

Iraqi Journal of Science

View full text Add to dashboard Cite

show abstract

“…Considering hierarchical class taxonomies in the classification problem has been widely studied in the literature [20][21][22][23][24]. The problem of hierarchical novelty detection actually comprises hierarchical classification of the known classes.…”

Section: Hierarchical Classificationmentioning

confidence: 99%

Hierarchical Novelty Detection for Traffic Sign Recognition

Ruiz

Serrat

2022

Sensors

View full text Add to dashboard Cite

Recent works have made significant progress in novelty detection, i.e., the problem of detecting samples of novel classes, never seen during training, while classifying those that belong to known classes. However, the only information this task provides about novel samples is that they are unknown. In this work, we leverage hierarchical taxonomies of classes to provide informative outputs for samples of novel classes. We predict their closest class in the taxonomy, i.e., its parent class. We address this problem, known as hierarchical novelty detection, by proposing a novel loss, namely Hierarchical Cosine Loss that is designed to learn class prototypes along with an embedding of discriminative features consistent with the taxonomy. We apply it to traffic sign recognition, where we predict the parent class semantics for new types of traffic signs. Our model beats state-of-the art approaches on two large scale traffic sign benchmarks, Mapillary Traffic Sign Dataset (MTSD) and Tsinghua-Tencent 100K (TT100K), and performs similarly on natural images benchmarks (AWA2, CUB). For TT100K and MTSD, our approach is able to detect novel samples at the correct nodes of the hierarchy with 81% and 36% of accuracy, respectively, at 80% known class accuracy.

show abstract

“…With CoLabel-SMBL, we could reduce model parameters, since we use a single branch. We compare CoLabel-SMBL against CoLabel and HML [5] in Table 8. CoLabel-SMBL sacrifices accuracy with the reduced parameters.…”

Section: Collaborative Learningmentioning

confidence: 99%

Constructive Interpretability with CoLabel: Corroborative Integration, Complementary Features, and Collaborative Learning

Suprem¹,

Vaidya²,

Cherkadi³

et al. 2022

Preprint

View full text Add to dashboard Cite

Machine learning models with explainable predictions are increasingly sought after, especially for real-world, critical applications that require bias detection and risk mitigation. Inherent interpretability, where a model is designed from the ground-up to provide prediction explanations, provide more faithful insights to model performance. In this paper, we present CoLabel, a constructively interpretable model that provides explanations rooted in the ground truth. We evaluate CoLabel for vehicle feature extraction, since these are crucial for re-id traffic monitoring and management, tracking, and vehicle make-model recognition (VMMR). Each task is a critical application that requires interpretable models. By construction, CoLabel provides VMMR along with a composite of interpretable features. These interpretable features are based on interpretable annotations of the ground truth labels. CoLabel first performs corroborative integration to join multiple datasets that each have a subset of desired interpretable annotations. Then Co-Label uses decomposable branches to extract interpretable features and fuses them together for final predictions. During feature fusion, CoLabel performs harmonization of the complementary branches so that VMMR features can be projected to the same semantic space for classification. We show that CoLabel achieves superior accuracy to the state-of-the-art black-box models. Further, CoLabel provides explanations that help in self-diagnosing and verifying predictions.

show abstract

Revisiting the CompCars Dataset for Hierarchical Car Classification: New Annotations, Experiments, and Results

Cited by 12 publications

References 24 publications

Adapted CNN-SMOTE-BGMM Deep Learning Framework for Network Intrusion Detection using Unbalanced Dataset

Adapted CNN-SMOTE-BGMM Deep Learning Framework for Network Intrusion Detection using Unbalanced Dataset

Hierarchical Novelty Detection for Traffic Sign Recognition

Constructive Interpretability with CoLabel: Corroborative Integration, Complementary Features, and Collaborative Learning

Contact Info

Product

Resources

About