Episodic Training for Domain Generalization Li, Da; Zhang, Jianshu; Yang, Yongxin ...
2019 IEEE/CVF International Conference on Computer Vision (ICCV),
2019-Oct.
Conference Proceeding
Odprti dostop
Domain generalization (DG) is the challenging and topical problem of learning models that generalize to novel testing domains with different statistics than a set of known training domains. The ...simple approach of aggregating data from all source domains and training a single deep neural network end-to-end on all the data provides a surprisingly strong baseline that surpasses many prior published methods. In this paper we build on this strong baseline by designing an episodic training procedure that trains a single deep network in a way that exposes it to the domain shift that characterises a novel domain at runtime. Specifically, we decompose a deep network into feature extractor and classifier components, and then train each component by simulating it interacting with a partner who is badly tuned for the current domain. This makes both components more robust, ultimately leading to our networks producing state-of-the-art performance on three DG benchmarks. Furthermore, we consider the pervasive workflow of using an ImageNet trained CNN as a fixed feature extractor for downstream recognition tasks. Using the Visual Decathlon benchmark, we demonstrate that our episodic-DG training improves the performance of such a general purpose feature extractor by explicitly training a feature for robustness to novel problems. This shows that DG training can benefit standard practice in computer vision.
Identification of non-diabetic renal disease (NDRD) in patients with type 2 diabetes mellitus (T2DM) may help tailor treatment. Intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) is a ...promising tool to evaluate renal function but its potential role in the clinical differentiation between diabetic nephropathy (DN) and NDRD remains unclear.
To investigate the added role of IVIM-DWI in the differential diagnosis between DN and NDRD in patients with T2DM.
Prospective.
Sixty-three patients with T2DM (ages: 22-69 years, 17 females) confirmed by renal biopsy divided into two subgroups (28 DN and 35 NDRD).
3 T/ T2 weighted imaging (T
WI), and intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI).
The parameters derived from IVIM-DWI (true diffusion coefficient D, pseudo-diffusion coefficient D*, and pseudo-diffusion fraction f) were calculated for the cortex and medulla, respectively. The clinical indexes related to renal function (eg cystatin C, etc.) and diabetes (eg diabetic retinopathy DR, fasting blood glucose, etc.) were measured and calculated within 1 week before MRI scanning. The clinical model based on clinical indexes and the IVIM-based model based on IVIM parameters and clinical indexes were established and evaluated, respectively.
Student's t-test; Mann-Whitney U test; Fisher's exact test; Chi-squared test; Intraclass correlation coefficient; Receiver operating characteristic analysis; Hosmer-Lemeshow test; DeLong's test. P < 0.05 was considered statistically significant.
The cortex D*, DR, and cystatin C values were identified as independent predictors of NDRD in multivariable analysis. The IVIM-based model, comprising DR, cystatin C, and cortex D*, significantly outperformed the clinical model containing only DR, and cystatin C (AUC = 0.934, 0.845, respectively).
The IVIM parameters, especially the renal cortex D* value, might serve as novel indicators in the differential diagnosis between DN and NDRD in patients with T2DM.
2 TECHNICAL EFFICACY: Stage 2.
Human free-hand sketches provide the useful data for studying human perceptual grouping, where the grouping principles such as the Gestalt laws of grouping are naturally in play during both the ...perception and sketching stages. In this paper, we make the first attempt to develop a universal sketch perceptual grouper. That is, a grouper that can be applied to sketches of any category created with any drawing style and ability, to group constituent strokes/segments into semantically meaningful object parts. The first obstacle to achieving this goal is the lack of large-scale datasets with grouping annotation. To overcome this, we contribute the largest sketch perceptual grouping dataset to date, consisting of 20 000 unique sketches evenly distributed over 25 object categories. Furthermore, we propose a novel deep perceptual grouping model learned with both generative and discriminative losses. The generative loss improves the generalization ability of the model, while the discriminative loss guarantees both local and global grouping consistency. Extensive experiments demonstrate that the proposed grouper significantly outperforms the state-of-the-art competitors. In addition, we show that our grouper is useful for a number of sketch analysis tasks, including sketch semantic segmentation, synthesis, and fine-grained sketch-based image retrieval.
Histone 3 Lys 27 trimethylation (H3K27me3)-mediated epigenetic silencing plays a critical role in multiple biological processes. However, the H3K27me3 recognition and transcriptional repression ...mechanisms are only partially understood. Here, we report a mechanism for H3K27me3 recognition and transcriptional repression. Our structural and biochemical data showed that the BAH domain protein AIPP3 and the PHD proteins AIPP2 and PAIPP2 cooperate to read H3K27me3 and unmodified H3K4 histone marks, respectively, in Arabidopsis. The BAH-PHD bivalent histone reader complex silences a substantial subset of H3K27me3-enriched loci, including a number of development and stress response-related genes such as the RNA silencing effector gene ARGONAUTE 5 (AGO5). We found that the BAH-PHD module associates with CPL2, a plant-specific Pol II carboxyl terminal domain (CTD) phosphatase, to form the BAH-PHD-CPL2 complex (BPC) for transcriptional repression. The BPC complex represses transcription through CPL2-mediated CTD dephosphorylation, thereby causing inhibition of Pol II release from the transcriptional start site. Our work reveals a mechanism coupling H3K27me3 recognition with transcriptional repression through the alteration of Pol II phosphorylation states, thereby contributing to our understanding of the mechanism of H3K27me3-dependent silencing.
In this paper, for the first time, we investigate the problem of generating 3D shapes from professional 2D sketches via deep learning. We target sketches done by professional artists, as these ...sketches are likely to contain more details than the ones produced by novices, and thus the reconstruction from such sketches poses a higher demand on the level of detail in the reconstructed models. This is importantly different to previous work, where the training and testing was conducted on either synthetic sketches or sketches done by novices. Novices sketches often depict shapes that are physically unrealistic, while models trained with synthetic sketches could not cope with the level of abstraction and style found in real sketches. To address this problem, we collected the first large-scale dataset of professional sketches, where each sketch is paired with a reference 3D shape, with a total of 1,500 professional sketches collected across 500 3D shapes. The dataset is available at http://sketchx.ai/downloads/ . We introduce two bespoke designs within a deep adversarial network to tackle the imprecision of human sketches and the unique figure/ground ambiguity problem inherent to sketch-based reconstruction. We show that existing 3D shapes generation methods designed for images fail to be naively applied to our problem, and demonstrate the effectiveness of our method both qualitatively and quantitatively.
We study the problem of fine-grained sketch-based image retrieval. By performing instance-level (rather than category-level) retrieval, it embodies a timely and practical application, particularly ...with the ubiquitous availability of touchscreens. Three factors contribute to the challenging nature of the problem: 1) free-hand sketches are inherently abstract and iconic, making visual comparisons with photos difficult; 2) sketches and photos are in two different visual domains, i.e., black and white lines versus color pixels; and 3) fine-grained distinctions are especially challenging when executed across domain and abstraction-level. To address these challenges, we propose to bridge the image-sketch gap both at the high level via parts and attributes, as well as at the low level via introducing a new domain alignment method. More specifically, first, we contribute a data set with 304 photos and 912 sketches, where each sketch and image is annotated with its semantic parts and associated part-level attributes. With the help of this data set, second, we investigate how strongly supervised deformable part-based models can be learned that subsequently enable automatic detection of part-level attributes, and provide pose-aligned sketch-image comparisons. To reduce the sketch-image gap when comparing low-level features, third, we also propose a novel method for instance-level domain-alignment that exploits both subspace and instance-level cues to better align the domains. Finally, fourth, these are combined in a matching framework integrating aligned low-level features, mid-level geometric structure, and high-level semantic attributes. Extensive experiments conducted on our new data set demonstrate effectiveness of the proposed method.
Sketch-based image retrieval (SBIR) is a challenging task due to the ambiguity inherent in sketches when compared with photos. In this paper, we propose a novel convolutional neural network based on ...Siamese network for SBIR. The main idea is to pull output feature vectors closer for input sketch-image pairs that are labeled as similar, and push them away if irrelevant. This is achieved by jointly tuning two convolutional neural networks which linked by one loss function. Experimental results on Flickr15K demonstrate that the proposed method offers a better performance when compared with several state-of-the-art approaches.
As a co-transcriptional process, RNA processing, including alternative splicing and alternative polyadenylation, is crucial for the generation of multiple mRNA isoforms. RNA processing mechanisms are ...widespread across all higher eukaryotes and play critical roles in cell differentiation, organ development and disease response. Recently, significant progresses have been made in understanding the mechanism of RNA processing. RNA processing is regulated by
trans
-acting factors such as splicing factors, RNA-binding proteins and
cis
-sequences in pre-mRNA, and increasing evidence suggests that epigenetic mechanisms, which are important for the dynamic regulation and state of specific chromatic regions, are also involved in co-transcriptional RNA processing. In contrast, recent studies also suggest that alternative RNA processing also has a feedback regulation on epigenetic mechanisms. In this review, we discuss recent studies and summarize the current knowledge on the epigenetic regulation of alternative RNA processing. In addition, a feedback regulation of RNA processing on epigenetic regulators is also discussed.
Extracellular vesicles (EVs) including exosomes can serve as mediators of cell-cell communication under physiological and pathological conditions. However, cargo molecules carried by EVs to exert ...their functions, as well as mechanisms for their regulated release and intake, have been poorly understood. In this study, we examined the effects of endothelial cells-derived EVs on neurons suffering from oxygen-glucose deprivation (OGD), which mimics neuronal ischemia-reperfusion injury in human diseases. In a human umbilical endothelial cell (HUVEC)-neuron coculture assay, we found that HUVECs reduced apoptosis of neurons under OGD, and this effect was compromised by GW4869, a blocker of exosome release. Purified EVs could be internalized by neurons and alleviate neuronal apoptosis under OGD. A miRNA, miR-1290, was highly enriched in HUVECs-derived EVs and was responsible for EV-mediated neuronal protection under OGD. Interestingly, we found that OGD enhanced intake of EVs by neurons cultured in vitro. We examined the expression of several potential receptors for EV intake and found that caveolin-1 (Cav-1) was upregulated in OGD-treated neurons and mice suffering from middle cerebral artery occlusion (MCAO). Knock-down of Cav-1 in neurons reduced EV intake, and canceled EV-mediated neuronal protection under OGD. HUVEC-derived EVs alleviated MCAO-induced neuronal apoptosis in vivo. These findings suggested that ischemia likely upregulates Cav-1 expression in neurons to increase EV intake, which protects neurons by attenuating apoptosis via miR-1290.
The multiple metastable excited states provided by excited‐state intramolecular proton transfer (ESIPT) molecules are beneficial to bring temperature‐dependent and color‐tunable long persistent ...luminescence (LPL). Meanwhile, ESIPT molecules are intrinsically suitable to be modulated as D‐π‐A structure to obtain both one/two‐photon excitation and LPL emission simultaneously. Herein, we report the rational design of a dynamic CdII coordination polymer (LIFM‐106) from ESIPT ligand to achieve the above goals. By comparing LIFM‐106 with the counterparts, we established a temperature‐regulated competitive relationship between singlet excimer and triplet LPL emission. The optimization of ligand aggregation mode effectively boost the competitiveness of the latter. In result, LIFM‐106 shows outstanding one/two‐photon excited LPL performance with wide temperature range (100–380 K) and tunable color (green to red). The multichannel radiation process was further elucidated by transient absorption and theoretical calculations, benefiting for the application in anti‐counterfeiting systems.
One/two‐photon‐excited long persistent luminescence (LPL) was obtained in an ESIPT‐attributed CdII coordination polymer (LIFM‐106) with wide temperature range (100–380 K) and color tunability. Comparative study manifests the rational ligand modification and J‐aggregation in LIFM‐106 enhances the competitiveness of monomer LPL to counter‐balance the excimer emission.