Exploring Simple and Transferable Recognition-Aware Image Processing.

Recent progress in image recognition has stimulated the deployment of vision systems at an unprecedented scale. As a result, visual data are now often consumed not only by humans but also by machines. Existing image processing methods only optimize for better human perception, yet the resulting images may not be accurately recognized by machines. This can be undesirable, e.g., the images can be improperly handled by search engines or recommendation systems. In this work, we examine simple approaches to improve machine recognition of processed images: optimizing the recognition loss directly on the image processing network or through an intermediate input transformation model. Interestingly, the processing model's ability to enhance recognition quality can transfer when evaluated on models of different architectures, recognized categories, tasks, and training datasets. This makes the methods applicable even when we do not have the knowledge of future recognition models, e.g., when uploading processed images to the Internet. We conduct experiments on multiple image processing tasks paired with ImageNet classification and PASCAL VOC detection as recognition tasks. With these simple yet effective methods, substantial accuracy gain can be achieved with strong transferability and minimal image quality loss. Through a user study we further show that the accuracy gain can transfer to a black-box cloud model. Finally, we try to explain this transferability phenomenon by demonstrating the similarities of different models' decision boundaries. Code is available at https://github.com/liuzhuang13/Transferable_RA.

Multi-View Supervision for Single-View Reconstruction via Differentiable Ray Consistency.

We study the notion of consistency between a 3D shape and a 2D observation and propose a differentiable formulation which allows computing gradients of the 3D shape given an observation from an arbitrary view. We do so by reformulating view consistency using a differentiable ray consistency (DRC) term. We show that this formulation can be incorporated in a learning framework to leverage different types of multi-view observations e.g., foreground masks, depth, color images, semantics etc. as supervision for learning single-view 3D prediction. We present empirical analysis of our technique in a controlled setting. We also show that this approach allows us to improve over existing techniques for single-view reconstruction of objects from the PASCAL VOC dataset.

Study on the Salivary Microbial Alteration of Men With Head and Neck Cancer and Its Relationship With Symptoms in Southwest China.

This study explored the association between oral microbes and head and neck cancer (HNC) as well as symptoms related to patients with HNC before surgical treatment. Fifty-six patients with HNC and 64 matched healthy controls were recruited from West China hospital in Southwest China. The demographic, clinical, and symptom data were collected. Salivary samples were collected to determine the microbial characteristics using 16S rRNA gene sequencing. Patients with HNC presented increased Capnocytophaga abundances. The oral microbial markers as Capnocytophaga (area under the curve=0.81) achieved a high classification power between the HNC patients and healthy controls. Moreover, using Capnocytophaga in conjunction with symptom of voice/speech difficulty achieved an overall predicting accuracy of 92.5% comparing with using Capnocytophaga alone (79.2% accuracy) in distinguishing the HNC patients from healthy controls. Salivary microbial profiles and HNC symptoms may be potential biomarkers for HNC screening.

[Distribution of chemical constituents in Changium smyrnioides].

OBJECTIVE: To analyze the distribution of chemical constituents in different parts of Changium smyrnioides. METHOD: The content of polysaccharides, mannitol and choline was determined by colorimetry and the water-soluble components was determined by HPLC. similarity coefficient of water-soluble components was calculated by the included angle cosine method. RESULT: The Contents of chemical constituents were significantly different between the leaves and the roots. The contents of choline and polysaccharides in the roots was higher than those in the leaves, but the contents of water-soluble extract and physiological ash in the leaves was higher than those in the roots. Contents of chemical constituents were also significantly different at different segments in the root, which decreased from the upper to the lower parts, increased from the outside to the inside parts except physiological ash, and the difference in horizontal segments was higher than that in vertical segments. Total peak area of water-soluble components in the leaves was 2.81 times higher than that in the roots, and gradually increased from the outside to the inside parts. CONCLUSION: Total content of chemical constituents in the leaves was significantly higher than that in the roots. The quality of medicinal used parts decreased gradually from the inside to the outside parts and from the upper to the lower parts in the root of Ch. smyrnioides.

[Analysis on water-soluble components in roots of Changium smyrnioides among different populations by HPLC].

OBJECTIVE: To analyze water-soluble components in the roots of Ch. smyrnioides among different populations that distributed in the main areas and give a reference for germplasm evaluation and quality control. METHOD: Water-soluble components were extracted with the cold-soaking method and analyzed by HPLC, similarity coefficient was calculated by included angle cosine method according to relative content of major water-soluble components, and systematic relationships were constructed based on UPGMA method. RESULT: There was significant difference in water-soluble components in root among population. Jiuhuashan population had the highest content of water-soluble extract. The content of water-soluble extract was below the pharmacopoeia standard in the root of Dalongshan population and Fushan population. There was significant difference in the HPLC chromatogram of water-soluble components in the root of Ch. smyrnioides from different populations, and the number of common peak was small. Similarity coefficient significantly ranged from 0.0306 to 0.9995 among 10 populations of Ch. smymrnioides. Water-soluble components in the root of Zijinshan population was the most unique, similarity coefficients were relatively small among Zijinshan population and the other seven populations except Hongshan population, and similarity coefficient was in a higher level of 0.9697 between Zijinshan population and Hongshan population. Water-soluble components were extremely similar in four populations that were Laoshan, Maoshan, Qinglongshan and Langyashan, and similarity coefficients among them were in a high level exceeded 0.99. 10 populations were divided into 3 groups according to clustering results. CONCLUSION: Water-soluble components show a high diversity in the roots of Ch. smyrnioides among different populations, and can be clearly divided into 3 types.

[Change of chemical constituents in Changium smyrnioides at different ages].

OBJECTIVE: To analyze the dynamic change of chemical constituents in Changium smyrnioides at different ages and give a reference for standard cultivation. METHOD: The contents of polysaccharides, mannitol and choline in the roots were determined by colorimetry. HPLC was used to analyze water-soluble components in the roots. GC-MS was used to analyze volatile oil components in the leaves. Similarity coefficient of water-soluble components was calculated by included angle cosine method. RESULT: There was significant difference in content of the analyzed chemical constituents except polysaccharide. Contents of mannitol, physiological ash and polysaccharide in the adult plant were higher than those in the young plant. On the contrast, content of water-soluble components and choline in the young plant was higher than that in the adult plant. With the increase of the growth years, water-soluble components in the roots and volatile oil components in the leaves were gradually concentrated. CONCLUSION: The growth time significantly affects the chemical constituents in Ch. smyrnioides, the reasonable harvest time is from the second to the forth year after transplanting.