The key to solving fine-grained image categorization is finding discriminate and local regions that correspond to subtle visual traits. Great strides have been made, with complex networks designed specifically to learn part-level discriminate feature representations. In this paper, we show that it is possible to cultivate subtle details without the need for overly complicated network designs or training mechanisms – a single loss is all it takes. The main trick lies with how we delve into individual feature channels early on, as opposed to the convention of starting from a consolidated feature map. The proposed loss function, termed as mutual-channel loss (MC-Loss), consists of two channel-specific components: a discriminality component and a diversity component. The discriminality component forces all feature channels belonging to the same class to be discriminative, through a novel channel-wise attention mechanism. The diversity component additionally constraints channels so that they become mutually exclusive across the spatial dimension. The end result is therefore a set of feature channels, each of which reflects different locally discriminative regions for a specific class. The MC-Loss can be trained end-to-end, without the need for any bounding-box/part annotations, and yields highly discriminative regions during inference. Experimental results show our MC-Loss when implemented on top of common base networks can achieve state-of-the-art performance on all four fine-grained categorization datasets (CUB-Birds, FGVC-Aircraft, Flowers-102, and Stanford Cars). Ablative studies further demonstrate the superiority of the MC-Loss when compared with other recently proposed general-purpose losses for visual classification, on two different base networks. Codes are available at: https://github.com/dongliangchang/Mutual-Channel-Loss .

The Devil is in the Channels: Mutual-Channel Loss for Fine-Grained Image Classification

Fine-grained visual classification (FGVC) is much more challenging than traditional classification tasks due to the inherently subtle intra-class object variations. Recent works are mainly part-driven (either explicitly or implicitly), with the assumption that fine-grained information naturally rests within the parts. In this paper, we take a different stance, and show that part operations are not strictly necessary – the key lies with encouraging the network to learn at different granularities and progressively fusing multi-granularity features together. In particular, we propose: (i) a progressive training strategy that effectively fuses features from different granularities, and (ii) a random jigsaw patch generator that encourages the network to learn features at specific granularities. We evaluate on several standard FGVC benchmark datasets, and show the proposed method consistently outperforms existing alternatives or delivers competitive results. The code is available at https://github.com/PRIS-CV/PMG-Progressive-Multi-Granularity-Training.

Fine-Grained Visual Classification via Progressive Multi-granularity Training of Jigsaw Patches

Binary cross entropy with deep learning technique for Image classification

Convolutional neural networks (CNNs) have recently shown excellent performance on the task of fine-grained vehicle classification, where the motivation is to identify the fine-grained categories of the given vehicles. Generally speaking, the main motivation of the conventional back-propagation algorithm is to optimize the loss function. The algorithm itself does not guarantee if the extracted features are discriminative for the task of classification. Intuitively, if we can learn more discriminative features with a relatively small number of feature maps, the generalization ability of the CNNs will be significantly improved. Therefore, we propose a channel max pooling (CMP) scheme, where a new layer is inserted between the fully connected layers and the convolutional layers. The proposed CMP scheme divides the feature maps into to several sub-groups. Then, it compresses the feature maps within each sub-group into a new one. The compression is carried out by selecting the maximum value among the same locations from different feature maps. Moreover, the proposed CMP layer has the advantage that it can reduce the number of parameters via reducing the number of channels in the CNNs. Experimental results on two fine-grained vehicle datasets demonstrate that the CMP modified CNNs can improve the classification accuracies on the task of fine-grained vehicle classification while a massive amount of parameters are reduced. Moreover, it has competitive performance when comparing with the-state-of-the-art methods.

Fine-Grained Vehicle Classification With Channel Max Pooling Modified CNNs

Key for solving fine-grained image categorization is finding discriminate and local regions that correspond to subtle visual traits. Great strides have been made, with complex networks designed specifically to learn part-level discriminate feature representations. In this paper, we show it is possible to cultivate subtle details without the need for overly complicated network designs or training mechanisms -- a single loss is all it takes. The main trick lies with how we delve into individual feature channels early on, as opposed to the convention of starting from a consolidated feature map. The proposed loss function, termed as mutual-channel loss (MC-Loss), consists of two channel-specific components: a discriminality component and a diversity component. The discriminality component forces all feature channels belonging to the same class to be discriminative, through a novel channel-wise attention mechanism. The diversity component additionally constraints channels so that they become mutually exclusive on spatial-wise. The end result is therefore a set of feature channels that each reflects different locally discriminative regions for a specific class. The MC-Loss can be trained end-to-end, without the need for any bounding-box/part annotations, and yields highly discriminative regions during inference. Experimental results show our MC-Loss when implemented on top of common base networks can achieve state-of-the-art performance on all four fine-grained categorization datasets (CUB-Birds, FGVC-Aircraft, Flowers-102, and Stanford-Cars). Ablative studies further demonstrate the superiority of MC-Loss when compared with other recently proposed general-purpose losses for visual classification, on two different base networks. Code available at this https URL

Fine-grained vehicle classification is a challenging topic in computer vision due to the high intraclass variance and low interclass variance. Recently, considerable progress has been made in fine-grained vehicle classification due to the huge success of deep neural networks. Most studies of fine-grained vehicle classification based on neural networks, focus on the neural network structure to improve the classification performance. In contrast to existing works on fine-grained vehicle classification, we focus on the loss function of the neural network. We add a regularization term to the cross-entropy loss and propose a new loss function, Dual Cross-Entropy Loss . The regularization term places a constraint on the probability that a data point is assigned to a class other than its ground-truth class, which can alleviate the vanishing of the gradient when the value of the cross-entropy loss is close to zero. To demonstrate the effectiveness of our loss function, we perform two sets of experiments. The first set is conducted on a small-sample fine-grained vehicle classification dataset, the Stanford Cars-196 dataset. The second set is conducted on two small-sample datasets, the LabelMe dataset and the UIUC-Sports dataset, as well as on one large-sample dataset, the CIFAR-10 dataset. The experimental results show that the proposed loss function improves the fine-grained vehicle classification performance and has good performance on three other general image classification tasks.

Dual Cross-Entropy Loss for Small-Sample Fine-Grained Vehicle Classification

Few-shot learning for fine-grained image classification has gained recent attention in computer vision. Among the approaches for few-shot learning, due to the simplicity and effectiveness, metric-based methods are favorably state-of-the-art on many tasks. Most of the metric-based methods assume a single similarity measure and thus obtain a single feature space. However, if samples can simultaneously be well classified via two distinct similarity measures, the samples within a class can distribute more compactly in a smaller feature space, producing more discriminative feature maps. Motivated by this, we propose a so-called Bi-Similarity Network ( BSNet ) that consists of a single embedding module and a bi-similarity module of two similarity measures. After the support images and the query images pass through the convolution-based embedding module, the bi-similarity module learns feature maps according to two similarity measures of diverse characteristics. In this way, the model is enabled to learn more discriminative and less similarity-biased features from few shots of fine-grained images, such that the model generalization ability can be significantly improved. Through extensive experiments by slightly modifying established metric/similarity based networks, we show that the proposed approach produces a substantial improvement on several fine-grained image benchmark datasets. Codes are available at: https://github.com/PRIS-CV/BSNet .

BSNet: Bi-Similarity Network for Few-shot Fine-grained Image Classification

Few-shot learning for fine-grained image classification has gained recent attention in computer vision. Among the approaches for few-shot learning, due to the simplicity and effectiveness, metric-based methods are favorably state-of-the-art on many tasks. Most of the metric-based methods assume a single similarity measure and thus obtain a single feature space. However, if samples can simultaneously be well classified via two distinct similarity measures, the samples within a class can distribute more compactly in a smaller feature space, producing more discriminative feature maps. Motivated by this, we propose a so-called \textit{Bi-Similarity Network} (\textit{BSNet}) that consists of a single embedding module and a bi-similarity module of two similarity measures. After the support images and the query images pass through the convolution-based embedding module, the bi-similarity module learns feature maps according to two similarity measures of diverse characteristics. In this way, the model is enabled to learn more discriminative and less similarity-biased features from few shots of fine-grained images, such that the model generalization ability can be significantly improved. Through extensive experiments by slightly modifying established metric/similarity based networks, we show that the proposed approach produces a substantial improvement on several fine-grained image benchmark datasets. Codes are available at: this https URL

Considering that in neural network based on softmax cross entropy loss, the output probability is mainly based on linear computation of parameter vectors of each class in the last layer and hidden features in the layer of sample points. Therefore, the final output of neural network is effected by of the L2-norm of parameter vector of each class. Taking binary-class as an example, if the parameter vector of a class has a large L2-norm., decision boundary is close to another class with smaller L2-norm., so that sample points will be easily assigned to the class with large L2-norm. Based on it, this paper proposes a new softmax cross entropy loss, which adjusts the position of decision boundary so that it is not biased to any class. Experimental results on the LabelMe dataset and the UIUC-Sports dataset show that the proposed loss is superior to softmax cross entropy loss.

Softmax Cross Entropy Loss with Unbiased Decision Boundary for Image Classification

Softmax cross-entropy loss with L2 regularization is commonly adopted in the machine learning and neural network community. Considering that the traditional softmax cross-entropy loss simply focuses on fitting or classifying the training data accurately but does not explicitly encourage a large decision margin for classification, some loss functions are proposed to improve the generalization performance by solving the problem. However, these loss functions enhance the difficulty of model optimization. In addition, inspired by regularized logistic regression, where the regularized term is responsible for adjusting the width of decision margin, which can be seen as an approximation of support vector machine, we proposed a large-margin regularization method for softmax cross-entropy loss. The advantages of the proposed loss are twofold as follows: the first is the generalization performance improvement, and the second is easy optimization. The experimental results on three small-sample datasets show that our regularization method achieves good performance and outperforms the existing popular regularization methods of neural networks.

/pdf/large-margin-regularized-softmax-cross-entropy-loss-5cnbv6mrfm.pdf

Jie Cao

Papers

Dual Cross-Entropy Loss for Small-Sample Fine-Grained Vehicle Classification

BSNet: Bi-Similarity Network for Few-shot Fine-grained Image Classification

BSNet: Bi-Similarity Network for Few-shot Fine-grained Image Classification

Softmax Cross Entropy Loss with Unbiased Decision Boundary for Image Classification

Large-Margin Regularized Softmax Cross-Entropy Loss