3D IoU-Net: IoU Guided 3D Object Detector for Point Clouds

Most existing point cloud based 3D object detectors focus on the tasks of classification and box regression. However, another bottleneck in this area is achieving an accurate detection confidence for the Non-Maximum Suppression (NMS) post-processing. In this paper, we add a 3D IoU prediction branch to the regular classification and regression branches. The predicted IoU is used as the detection confidence for NMS. In order to obtain a more accurate IoU prediction, we propose a 3D IoU-Net with IoU sensitive feature learning and an IoU alignment operation. To obtain a perspective-invariant prediction head, we propose an Attentive Corner Aggregation (ACA) module by aggregating a local point cloud feature from each perspective of eight corners and adaptively weighting the contribution of each perspective with different attentions. We propose a Corner Geometry Encoding (CGE) module for geometry information embedding. To the best of our knowledge, this is the first time geometric embedding information has been introduced in proposal feature learning. These two feature parts are then adaptively fused by a multi-layer perceptron (MLP) network as our IoU sensitive feature. The IoU alignment operation is introduced to resolve the mismatching between the bounding box regression head and IoU prediction, thereby further enhancing the accuracy of IoU prediction. The experimental results on the KITTI car detection benchmark show that 3D IoU-Net with IoU perception achieves state-of-the-art performance.

FAIRS -- Soft Focus Generator and Attention for Robust Object Segmentation from Extreme Points

Semantic segmentation from user inputs has been actively studied to facilitate interactive segmentation for data annotation and other applications. Recent studies have shown that extreme points can be effectively used to encode user inputs. A heat map generated from the extreme points can be appended to the RGB image and input to the model for training. In this study, we present FAIRS -- a new approach to generate object segmentation from user inputs in the form of extreme points and corrective clicks. We propose a novel approach for effectively encoding the user input from extreme points and corrective clicks, in a novel and scalable manner that allows the network to work with a variable number of clicks, including corrective clicks for output refinement. We also integrate a dual attention module with our approach to increase the efficacy of the model in preferentially attending to the objects. We demonstrate that these additions help achieve significant improvements over state-of-the-art in dense object segmentation from user inputs, on multiple large-scale datasets. Through experiments, we demonstrate our method's ability to generate high-quality training data as well as its scalability in incorporating extreme points, guiding clicks, and corrective clicks in a principled manner.

Pathological Retinal Region Segmentation From OCT Images Using Geometric Relation Based Augmentation

Medical image segmentation is an important task for computer aided diagnosis. Pixelwise manual annotations of large datasets require high expertise and is time consuming. Conventional data augmentations have limited benefit by not fully representing the underlying distribution of the training set, thus affecting model robustness when tested on images captured from different sources. Prior work leverages synthetic images for data augmentation ignoring the interleaved geometric relationship between different anatomical labels. We propose improvements over previous GAN-based medical image synthesis methods by jointly encoding the intrinsic relationship of geometry and shape. Latent space variable sampling results in diverse generated images from a base image and improves robustness. Given those augmented images generated by our method, we train the segmentation network to enhance the segmentation performance of retinal optical coherence tomography (OCT) images. The proposed method outperforms state-of-the-art segmentation methods on the public RETOUCH dataset having images captured from different acquisition procedures. Ablation studies and visual analysis also demonstrate benefits of integrating geometry and diversity.

Controllable Orthogonalization in Training DNNs

Orthogonality is widely used for training deep neural networks (DNNs) due to its ability to maintain all singular values of the Jacobian close to 1 and reduce redundancy in representation. This paper proposes a computationally efficient and numerically stable orthogonalization method using Newton's iteration (ONI), to learn a layer-wise orthogonal weight matrix in DNNs. ONI works by iteratively stretching the singular values of a weight matrix towards 1. This property enables it to control the orthogonality of a weight matrix by its number of iterations. We show that our method improves the performance of image classification networks by effectively controlling the orthogonality to provide an optimal tradeoff between optimization benefits and representational capacity reduction. We also show that ONI stabilizes the training of generative adversarial networks (GANs) by maintaining the Lipschitz continuity of a network, similar to spectral normalization (SN), and further outperforms SN by providing controllable orthogonality.

Architecture Disentanglement for Deep Neural Networks

Deep Neural Networks (DNNs) are central to deep learning, and understanding their internal working mechanism is crucial if they are to be used for emerging applications in medical and industrial AI. To this end, the current line of research typically involves linking semantic concepts to a DNN's units or layers. However, this fails to capture the hierarchical inference procedure throughout the network. To address this issue, we introduce the novel concept of Neural Architecture Disentanglement (NAD) in this paper. Specifically, we disentangle a pre-trained network into hierarchical paths corresponding to specific concepts, forming the concept feature paths, i.e., the concept flows from the bottom to top layers of a DNN. Such paths further enable us to quantify the interpretability of DNNs according to the learned diversity of human concepts. We select four types of representative architectures ranging from handcrafted to autoML-based, and conduct extensive experiments on object-based and scene-based datasets. Our NAD sheds important light on the information flow of semantic concepts in DNNs, and provides a fundamental metric that will facilitate the design of interpretable network architectures. Code will be available at: https://github.com/hujiecpp/NAD.

An Investigation into the Stochasticity of Batch Whitening

Batch Normalization (BN) is extensively employed in various network architectures by performing standardization within mini-batches. A full understanding of the process has been a central target in the deep learning communities. Unlike existing works, which usually only analyze the standardization operation, this paper investigates the more general Batch Whitening (BW). Our work originates from the observation that while various whitening transformations equivalently improve the conditioning, they show significantly different behaviors in discriminative scenarios and training Generative Adversarial Networks (GANs). We attribute this phenomenon to the stochasticity that BW introduces. We quantitatively investigate the stochasticity of different whitening transformations and show that it correlates well with the optimization behaviors during training. We also investigate how stochasticity relates to the estimation of population statistics during inference. Based on our analysis, we provide a framework for designing and comparing BW algorithms in different scenarios. Our proposed BW algorithm improves the residual networks by a significant margin on ImageNet classification. Besides, we show that the stochasticity of BW can improve the GAN's performance with, however, the sacrifice of the training stability.

Pedestrian Detection: The Elephant In The Room

Pedestrian detection is used in many vision based applications ranging from video surveillance to autonomous driving. Despite achieving high performance, it is still largely unknown how well existing detectors generalize to unseen data. To this end, we conduct a comprehensive study in this paper, using a general principle of direct cross-dataset evaluation. Through this study, we find that existing state-of-the-art pedestrian detectors generalize poorly from one dataset to another. We demonstrate that there are two reasons for this trend. Firstly, they over-fit on popular datasets in a traditional single-dataset training and test pipeline. Secondly, the training source is generally not dense in pedestrians and diverse in scenarios. Accordingly, through experiments we find that a general purpose object detector works better in direct cross-dataset evaluation compared with state-of-the-art pedestrian detectors and we illustrate that diverse and dense datasets, collected by crawling the web, serve to be an efficient source of pre-training for pedestrian detection. Furthermore, we find that a progressive training pipeline works good for autonomous driving oriented detector. We improve upon previous state-of-the-art on reasonable/heavy subsets of CityPersons dataset by 1.3%/1.7% and on Caltech by 1.8%/14.9% in terms of log average miss rate (MR^2) points without any fine-tuning on the test set. Detector trained through proposed pipeline achieves top rank at the leaderborads of CityPersons [42] and ECP [4]. Code and models will be available at https://github.com/hasanirtiza/Pedestron.

Latent Embedding Feedback and Discriminative Features for Zero-Shot Classification

Zero-shot learning strives to classify unseen categories for which no data is available during training. In the generalized variant, the test samples can further belong to seen or unseen categories. The state-of-the-art relies on Generative Adversarial Networks that synthesize unseen class features by leveraging class-specific semantic embeddings. During training, they generate semantically consistent features, but discard this constraint during feature synthesis and classification. We propose to enforce semantic consistency at all stages of (generalized) zero-shot learning: training, feature synthesis and classification. We further introduce a feedback loop, from a semantic embedding decoder, that iteratively refines the generated features during both the training and feature synthesis stages. The synthesized features together with their corresponding latent embeddings from the decoder are transformed into discriminative features and utilized during classification to reduce ambiguities among categories. Experiments on (generalized) zero-shot learning for object and action classification reveal the benefit of semantic consistency and iterative feedback for GAN-based networks, outperforming existing methods on six zero-shot learning benchmarks.