Balancing the Encoder and Decoder Complexity in Image Compression for Classification.

This paper presents a study on the computational complexity of coding for machines, with a focus on image coding for classification. We first conduct a comprehensive set of experiments to analyze the size of the encoder (which encodes images to bitstreams), the size of the decoder (which decodes bitstreams and predicts class labels), and their impact on the rate-accuracy trade-off in compression for classification. Through empirical investigation, we demonstrate a complementary relationship between the encoder size and the decoder size, i.e., it is better to employ a large encoder with a small decoder and vice versa. Motivated by this relationship, we introduce a feature compression-based method for efficient image compression for classification. By compressing features at various layers of a neural network-based image classification model, our method achieves adjustable rate, accuracy, and encoder (or decoder) size using a single model. Experimental results on ImageNet classification show that our method achieves competitive results with existing methods while being much more flexible. The code will be made publicly available.

Integration of USDA Food Classification System and Food Composition Database for Image-Based Dietary Assessment among Individuals Using Insulin.

New imaging technologies to identify food can reduce the reporting burden of participants but heavily rely on the quality of the food image databases to which they are linked to accurately identify food images. The objective of this study was to develop methods to create a food image database based on the most commonly consumed U.S. foods and those contributing the most to energy. The objective included using a systematic classification structure for foods based on the standardized United States Department of Agriculture (USDA) What We Eat in America (WWEIA) food classification system that can ultimately be used to link food images to a nutrition composition database, the USDA Food and Nutrient Database for Dietary Studies (FNDDS). The food image database was built using images mined from the web that were fitted with bounding boxes, identified, annotated, and then organized according to classifications aligning with USDA WWEIA. The images were classified by food category and subcategory and then assigned a corresponding USDA food code within the USDA's FNDDS in order to systematically organize the food images and facilitate a linkage to nutrient composition. The resulting food image database can be used in food identification and dietary assessment.

Long-Tailed Food Classification.

Food classification serves as the basic step of image-based dietary assessment to predict the types of foods in each input image. However, foods in real-world scenarios are typically long-tail distributed, where a small number of food types are consumed more frequently than others, which causes a severe class imbalance issue and hinders the overall performance. In addition, none of the existing long-tailed classification methods focus on food data, which can be more challenging due to the inter-class similarity and intra-class diversity between food images. In this work, two new benchmark datasets for long-tailed food classification are introduced, including Food101-LT and VFN-LT, where the number of samples in VFN-LT exhibits real-world long-tailed food distribution. Then, a novel two-phase framework is proposed to address the problem of class imbalance by (1) undersampling the head classes to remove redundant samples along with maintaining the learned information through knowledge distillation and (2) oversampling the tail classes by performing visually aware data augmentation. By comparing our method with existing state-of-the-art long-tailed classification methods, we show the effectiveness of the proposed framework, which obtains the best performance on both Food101-LT and VFN-LT datasets. The results demonstrate the potential to apply the proposed method to related real-life applications.

AP-57/C10orf99 is a new type of multifunctional antimicrobial peptide.

Antimicrobial peptides (AMPs) are an evolutionarily conserved component of the innate immune response that provides host defence at skin and mucosal surfaces. Here, we report the identification and characterization of a new type human AMPs, termed AP-57 (Antimicrobial Peptide with 57 amino acid residues), which is also known as C10orf99 (chromosome 10 open reading frame 99). AP-57 is a short basic amphiphilic peptide with four cysteines and a net charge +14 (MW = 6.52, PI = 11.28). The highest expression of AP-57 were detected in the mucosa of stomach and colon through immunohistochemical assay. Epithelium of skin and esophagus show obvious positive staining and strong positive staining were also observed in some tumor and/or their adjacent tissues, such as esophagus cancer, hepatocellular carcinoma, squamous cell carcinoma and invasive ductal carcinoma. AP-57 exhibited broad-spectrum antimicrobial activities against Gram-positive Staphylococcus aureus, Actinomyce, and Fungi Aspergillus niger as well as mycoplasma and lentivirus. AP-57 also exhibited DNA binding capacity and specific cytotoxic effects against human B-cell lymphoma Raji. Compared with other human AMPs, AP-57 has its distinct characteristics, including longer sequence length, four cysteines, highly cationic character, cell-specific toxicity, DNA binding and tissue-specific expressing patterns. Together, AP-57 is a new type of multifunctional AMPs worthy further investigation.