Deep neural networks can model images with rich latent representations, but they cannot naturally conceptualize structures of object categories in a human-perceptible way. This paper addresses the problem of learning object structures in an image modeling process without supervision. We propose an autoencoding formulation to discover landmarks as explicit structural representations. The encoding module outputs landmark coordinates, whose validity is ensured by constraints that reflect the necessary properties for landmarks. The decoding module takes the landmarks as a part of the learnable input representations in an end-to-end differentiable framework. Our discovered landmarks are semantically meaningful and more predictive of manually annotated landmarks than those discovered by previous methods. The coordinates of our landmarks are also complementary features to pretrained deep-neural-network representations in recognizing visual attributes. In addition, the proposed method naturally creates an unsupervised, perceptible interface to manipulate object shapes and decode images with controllable structures. The project web page: http://ytzhang.net/projects/lmdis-rep arXiv:1804.04412v1 [cs.CV] 12 Apr 2018 3. The discovered landmarks show strong discriminative performance in recognizing visual attributes. 4. Our landmark-based image decoder is useful for controllable image decoding, such as object shape manipulation and structure-conditioned image generation. 2. Related work Discriminative part learning. Parts are commonly used object structures in computer vision. The deformable partbased model [15] learns object part configurations to optimize the object detection accuracy, where similar ideas are rooted in earlier constellation approaches [16, 66, 6]. A recent method [72] based on the deep neural network performs end-to-end learning of deformable mixture of parts for pose estimation. The recurrent architecture [19] and spatial transformer network [23] are also used to discover and refine object parts for fine-grained image classification [27]. In addition, discriminative mid-level patches can be also discovered without explicit supervision [54]. Object-part discovery based on subspace analysis and clustering techniques is also shown to improve neural-network-based image recognition [52]. Unlike the approaches specific to discriminative tasks, our work focuses on learning landmarks for generic image modeling. Learning structural representations. To capture the intrinsic structures of objects, existing studies [44, 45, 37] disentangle visual content into multiple factors of variations, like the camera viewpoint, motion, and identity. The physical parameters of these factors are, however, still embedded in non-perceptible latent representations. Methods based on multi-task learning [78, 21, 65, 81] can take conceptualized structures (e.g., landmarks, masks, depth) as additional outputs. These structures in this setting are designed by humans and require supervision to learn. Learning explicit structures for image c...

No abstract

Recognition of viral RNA by the retinoic acid-inducible gene-I (RIG-I)-like receptors (RLRs) initiates innate antiviral immune response. How the binding of viral RNA to and activation of the RLRs are regulated remains enigmatic. In this study, we identified ZCCHC3 as a positive regulator of the RLRs including RIG-I and MDA5. ZCCHC3 deficiency markedly inhibited RNA virus-triggered induction of downstream antiviral genes, and ZCCHC3-deficient mice were more susceptible to RNA virus infection. ZCCHC3 was associated with RIG-I and MDA5 and functions in two distinct processes for regulation of RIG-I and MDA5 activities. ZCCHC3 bound to dsRNA and enhanced the binding of RIG-I and MDA5 to dsRNA. ZCCHC3 also recruited the E3 ubiquitin ligase TRIM25 to the RIG-I and MDA5 complexes to facilitate its K63-linked polyubiquitination and activation. Thus, ZCCHC3 is a co-receptor for RIG-I and MDA5, which is critical for RLR-mediated innate immune response to RNA virus.

We introduce a View-Volume convolutional neural network (VVNet) for inferring the occupancy and semantic labels of a volumetric 3D scene from a single depth image. The VVNet concatenates a 2D view CNN and a 3D volume CNN with a differentiable projection layer. Given a single RGBD image, our method extracts the detailed geometric features from the input depth image with a 2D view CNN and then projects the features into a 3D volume according to the input depth map via a projection layer. After that, we learn the 3D context information of the scene with a 3D volume CNN for computing the result volumetric occupancy and semantic labels. With combined 2D and 3D representations, the VVNet efficiently reduces the computational cost, enables feature extraction from multi-channel high resolution inputs, and thus significantly improves the result accuracy. We validate our method and demonstrate its efficiency and effectiveness on both synthetic SUNCG and real NYU dataset.