Recent medical imaging technologies, specifically functional magnetic resonance imaging (fMRI), have advanced the diagnosis of neurological and neurodevelopmental disorders by allowing scientists and physicians to observe the activity within and between different regions of the brain. Deep learning methods have frequently been implemented to analyze images produced by such technologies and perform disease classification tasks; however, current state-of-the-art approaches do not take advantage of all the information offered by fMRI scans. In this paper, we propose a deep multimodal model that learns a joint representation from two types of connectomic data offered by fMRI scans. Incorporating two functional imaging modalities in an automated end-to-end autism diagnosis system will offer a more comprehensive picture of the neural activity, and thus allow for more accurate diagnoses. Our multimodal training strategy achieves a classification accuracy of 74% and a recall of 95%, as well as an F1 score of 0.805, and its overall performance is superior to using only one type of functional data.

Implicit neural representations store videos as neural networks and have performed well for various vision tasks such as video compression and denoising. With frame index or positional index as input, implicit representations (NeRV, E-NeRV, \etc) reconstruct video from fixed and content-agnostic embeddings. Such embedding largely limits the regression capacity and internal generalization for video interpolation. In this paper, we propose a Hybrid Neural Representation for Videos (HNeRV), where a learnable encoder generates content-adaptive embeddings, which act as the decoder input. Besides the input embedding, we introduce HNeRV blocks, which ensure model parameters are evenly distributed across the entire network, such that higher layers (layers near the output) can have more capacity to store high-resolution content and video details. With content-adaptive embeddings and re-designed architecture, HNeRV outperforms implicit methods in video regression tasks for both reconstruction quality ($+4.7$ PSNR) and convergence speed ($16\times$ faster), and shows better internal generalization. As a simple and efficient video representation, HNeRV also shows decoding advantages for speed, flexibility, and deployment, compared to traditional codecs~(H.264, H.265) and learning-based compression methods. Finally, we explore the effectiveness of HNeRV on downstream tasks such as video compression and video inpainting. We provide project page at https://haochen-rye.github.io/HNeRV, and Code at https://github.com/haochen-rye/HNeRV

Incorporating relational reasoning in neural networks for object recognition remains an open problem. Although many attempts have been made for relational reasoning, they generally only consider a single type of relationship. For example, pixel relations through self-attention (e.g., non-local networks), scale relations through feature fusion (e.g., feature pyramid networks), or object relations through graph convolutions (e.g., reasoning-RCNN). Little attention has been given to more generalized frameworks that can reason across these relationships. In this paper, we propose a hierarchical relational reasoning framework (HR-RCNN) for object detection, which utilizes a novel graph attention module (GAM). This GAM is a concise module that enables reasoning across heterogeneous nodes by operating on the graph edges directly. Leveraging heterogeneous relationships, our HR-RCNN shows great improvement on COCO dataset, for both object detection and instance segmentation.

Ensemble learning is a general technique to improve accuracy in machine learning. However, the heavy computation of a ConvNets ensemble limits its usage in deep learning. In this paper, we present Group Ensemble Network (GENet), an architecture incorporating an ensemble of ConvNets in a single ConvNet. Through a shared-base and multi-head structure, GENet is divided into several groups to make explicit ensemble learning possible in a single ConvNet. Owing to group convolution and the shared-base, GENet can fully leverage the advantage of explicit ensemble learning while retaining the same computation as a single ConvNet. Additionally, we present Group Averaging, Group Wagging and Group Boosting as three different strategies to aggregate these ensemble members. Finally, GENet outperforms larger single networks, standard ensembles of smaller networks, and other recent state-of-the-art methods on CIFAR and ImageNet. Specifically, group ensemble reduces the top-1 error by 1.83% for ResNeXt-50 on ImageNet. We also demonstrate its effectiveness on action recognition and object detection tasks.

Recent development of object detection mainly depends on deep learning with large-scale benchmarks. However, collecting such fully-annotated data is often difficult or expensive for real-world applications, which restricts the power of deep neural networks in practice. Alternatively, humans can detect new objects with little annotation burden, since humans often use the prior knowledge to identify new objects with few elaborately-annotated examples, and subsequently generalize this capacity by exploiting objects from wild images. Inspired by this procedure of learning to detect, we propose a novel Progressive Object Transfer Detection (POTD) framework. Specifically, we make three main contributions in this paper. First, POTD can leverage various object supervision of different domains effectively into a progressive detection procedure. Via such human-like learning, one can boost a target detection task with few annotations. Second, POTD consists of two delicate transfer stages, i.e., Low-Shot Transfer Detection (LSTD), and Weakly-Supervised Transfer Detection (WSTD). In LSTD, we distill the implicit object knowledge of source detector to enhance target detector with few annotations. It can effectively warm up WSTD later on. In WSTD, we design a recurrent object labelling mechanism for learning to annotate weakly-labeled images. More importantly, we exploit the reliable object supervision from LSTD, which can further enhance the robustness of target detector in the WSTD stage. Finally, we perform extensive experiments on a number of challenging detection benchmarks with different settings. The results demonstrate that, our POTD outperforms the recent state-of-the-art approaches.