Forest fires are among the most critical natural tragedies threatening forest lands and resources. The accurate and early detection of forest fires is essential to reduce losses and improve firefighting. Conventional firefighting techniques, based on ground inspection and limited by the field-of-view, lead to insufficient monitoring capabilities for large areas. Recently, due to their excellent flexibility and ability to cover large regions, unmanned aerial vehicles (UAVs) have been used to combat forest fire incidents. An essential step for an autonomous system that monitors fire situations is first to locate the fire in a video. State-of-the-art forest-fire segmentation methods based on vision transformers (ViTs) and convolutional neural networks (CNNs) use a single aerial image. Nevertheless, fire has an inconsistent scale and form, and small fires from long-distance cameras lack salient features, so accurate fire segmentation from a single image has been challenging. In addition, the techniques based on CNNs treat all image pixels equally and overlook global information, limiting their performance, while ViT-based methods suffer from high computational overhead. To address these issues, we proposed a spatiotemporal architecture called FFS-UNet, which exploited temporal information for forest-fire segmentation by combining a transformer into a modified lightweight UNet model. First, we extracted a keyframe and two reference frames using three different encoder paths in parallel to obtain shallow features and perform feature fusion. Then, we used a transformer to perform deep temporal-feature extraction, which enhanced the feature learning of the fire pixels and made the feature extraction more robust. Finally, we combined the shallow features of the keyframe for de-convolution in the decoder path via skip-connections to segment the fire. We evaluated empirical outcomes on the UAV-collected video and Corsican Fire datasets. The proposed FFS-UNet demonstrated enhanced performance with fewer parameters by achieving an F1-score of 95.1% and an IoU of 86.8% on the UAV-collected video, and an F1-score of 91.4% and an IoU of 84.8% on the Corsican Fire dataset, which were higher than previous forest fire techniques. Therefore, the suggested FFS-UNet model effectively resolved fire-monitoring issues with UAVs.

Generative Adversarial Networks (GAN) is a popular machine learning framework that possesses powerful image generation ability, which is useful for different multimedia applications (e.g. photographic filters, image editing). However, typical GAN models have a large memory footprint that limits their practical applications for resource-constrained devices (e.g. smartphones). To deploy GAN models on these devices, we propose AdjustableGAN, a filter-wise evolutionary pruning algorithm, whichunlike previous workscan compress generative models to user-specified compression ratio while preserving the models performance. We achieve this by using a Genetic Algorithm (GA) to derive pruned models that are tuned with a balanced performance score and compression ratio. Additional GAs are utilized to intelligently adjust (i.e. upsize, downsize) the compression ratio of pruned models by finding optimal filter activation/deactivation. The upsizing operation is performed such that the restored filters have maximal performance increase. For downsizing, the pruned filters are ensured to have minimal impact on performance. Our experimental results show that our method can reliably compress generative networks with minimal accuracy drop compared to other state-of-the-art compression algorithms.

Sarcasm is an expression of contempt expressed through verbal irony. It is a nuanced form of language that individuals use to imply the opposite of what they are actually saying, and thus it can be difficult to detect at times. The lack of large, annotated datasets is one of the major challenges and limitations of building systems to detect sarcasm automatically. To address this issue, we propose a hybrid graph-based framework, namely, SINCERE, to build compact sarcasm detection models with sentiment and emotion analysis by leveraging only a small amount of prior data. To automatically extract patterns from a small dataset collected by distant supervision, a graph is first constructed. This approach is used to discover latent representations of vertices in a network, as the basis for a language model. We demonstrate that simple classifiers built from the model can detect sarcasm and generalize better than the state-of-the-art approach. According to the experimental results, the proposed SINCERE framework is able to outperform the SOTA baselines on accuracy by 5%.

In this article, we propose to study a novel research problem to boost group performance, that is, social-aware diversity-optimized group extraction (SDGE), which takes into consideration the two important factors: 1) group diversity and 2) social tightness. We prove the NP-hardness of SDGE and propose an effective algorithm, named group shrinking for diversity maximization (GSDM) with a performance guarantee, that is, GSDM is a three-approximation algorithm to the SDGE problem studied in this article. We further propose three effective pruning strategies that are able to boost the efficiency of GSDM but do not deteriorate its performance. We conduct extensive experiments on multiple large-scale real datasets to evaluate the performance of GSDM. The experimental results show that our proposed GSDM outperforms the other baseline approaches significantly, in terms of solution quality and efficiency. Moreover, the experimental results also confirm that our proposed pruning strategies indeed boost the efficiency of the algorithm.

Unpaired image-to-image translation finds a mapping between two domains that do not have paired data. One approach is patchwise contrastive learning, a one-sided translation that maximizes mutual information between corresponding input and output patches. Non-corresponding patches are treated as negatives. Previous approaches randomly select non-corresponding patches, resulting in semantically similar patches incorrectly labeled as negatives. Inspired by negative learning, we propose the novel patch-wise negative learning loss to address this issue. We do not naively minimize mutual information between all non-corresponding ones, unlike prior methods. Instead, we choose one non-corresponding patch and maximize dissimilarity with the query patch. The selected non-corresponding patch reduces the chance of choosing false negatives that contain high mutual information. By further maximizing dissimilarity with that single negative, we discourage our model from fitting on noisy negative patches. We demonstrate the capabilities of our model against other prominent image translation methods on the selfie2anime, horse2zebra, and cat2dog datasets.