Search

Learning to recognize novel visual classes from few samples is challenging but promising. Previous studies have shown that few-shot model tends to overfit and lead to poor generalization performance, which is because it finds a biased distribution based on a few samples. In addition, in agriculture-specific domains, there are more serious research challenges such as imbalanced disease distribution, one-shot representation biases, fine-grained recognition, and granularity shift. As far as we know, this study is the first work on the fine-grained "Coarse-to-Fine" few-shot plant disease classification, which classifies "fine-grained novel classes" (specific to disease severity) based on "coarse-grained base classes" (specific to plant species). A complete two-stage in-use calibration strategy is presented in this paper. Firstly, we propose an attention-based inverse Mahalanobis distance weighted prototype calibration module (AIPCM). By transferring statistics from sample-rich coarse-grained base classes to sample-scarce fine-grained novel classes, we achieve prototype calibration for 1-shot sample and obtain an unbiased distribution in the feature space. Secondly, to generate more reasonable decision boundaries, we propose a prior-driven task-adapted decision boundary calibration module (TDBCM) based on class-covariance metric. The original Euclidean/Cosine distance is updated to the Mahalanobis distance by introducing the prior mean and covariance of the high-dimensional features. Experimental results on several datasets demonstrate that our model outperforms the state-of-the-art (SOTA) models. It can be said that our work is a valuable supplement to the domain-specific agricultural applications. [ABSTRACT FROM AUTHOR]

In this paper, we explore multi-level semantic information of human body structure and propose a paradigm for bottom-up multi-person pose estimation. To represent the multi-level semantic body structure, we define a Spatial Hierarchical Body Tree (SHBT) that encodes the location and association information of the body center, parts, and joints for each human instance. This encoding approach assists in associating joints to each human instance, and the multi-level form is suitable for handling cases of partial human body occlusion. To apply the spatial hierarchical body tree to multi-person pose estimation, we build Hierarchical Pose Net(Heap-net) by inheriting the topology of the SHBT. This Heap-net explicitly defines the corresponding output order and the feature fusion aggregation. Furthermore, we propose a shared filters spatial pyramid module, which consists of a multi-branches dilation convolution module with shared filters and a max-out activation, to alleviate the effect of a wide range of human scale. To verify the effectiveness of our model, we conduct experiments on the MSCOCO keypoints detection validation and test set. The experimental results are comparable to the previous bottom-up multi-person pose estimation methods. [ABSTRACT FROM AUTHOR]

We present a framework for artwork image synthesis from unsupervised segmentation maps input and style images. The output has style consistency with style images and the semantic structure from the corresponding segmentation label. Existing methods of transferring semantic labels to painting images require large amounts of manual segmentation pairs for training. To address the issue, we use unsupervised segmentation maps to build on training pairs and learn the generator with the proposed spatially adaptive instance normalization block. Our method exploits the style consistency and semantic consistency loss functions to reduce the artifact in synthetic images. Extensive experiments in several image translation tasks show the effectiveness of our method in generating an image with both structures of segmentation and style of exemplar image. [ABSTRACT FROM AUTHOR]

Artistic image synthesis is receiving increasing engagement in the multimedia community because of the development and improvement of generative adversarial networks. Digital art synthesis methods perform uncontrolled manipulation in complicated landscape scenarios because of the domain diversity of the paintings. To solve this problem, this paper presents the tag-guided correlation matrix for matching the tag with the content code from the source image as better guidance on the style code. Correspondingly, the new generator module adapts the warped style code onto the semantic feature for controllable synthesis. Extensive experiments in several image synthesis tasks show the effectiveness of the proposed method in generating images with variations or multi-tag combinations. In addition, the proposed method shows better quantitative performance than recent conditional image synthesis approaches in artistic image manipulation tasks. [ABSTRACT FROM AUTHOR]