Recent advances in deep learning, especially deep convolutional neural networks (CNNs), have led to significant improvement over previous semantic segmentation systems. Here we show how to improve pixel-wise semantic segmentation by manipulating convolution-related operations that are of both theoretical and practical value. First, we design dense upsampling convolution (DUC) to generate pixel-level prediction, which is able to capture and decode more detailed information that is generally missing in bilinear upsampling. Second, we propose a hybrid dilated convolution (HDC) framework in the encoding phase. This framework 1) effectively enlarges the receptive fields (RF) of the network to aggregate global information; 2) alleviates what we call the "gridding issue"caused by the standard dilated convolution operation. We evaluate our approaches thoroughly on the Cityscapes dataset, and achieve a state-of-art result of 80.1% mIOU in the test set at the time of submission. We also have achieved state-of-theart overall on the KITTI road estimation benchmark and the PASCAL VOC2012 segmentation task. Our source code can be found at https://github.com/TuSimple/TuSimple-DUC.

Understanding Convolution for Semantic Segmentation

Full-image dependencies provide useful contextual information to benefit visual understanding problems. In this work, we propose a Criss-Cross Network (CCNet) for obtaining such contextual information in a more effective and efficient way. Concretely, for each pixel, a novel criss-cross attention module in CCNet harvests the contextual information of all the pixels on its criss-cross path. By taking a further recurrent operation, each pixel can finally capture the full-image dependencies from all pixels. Overall, CCNet is with the following merits: 1) GPU memory friendly. Compared with the non-local block, the proposed recurrent criss-cross attention module requires 11x less GPU memory usage. 2) High computational efficiency. The recurrent criss-cross attention significantly reduces FLOPs by about 85% of the non-local block in computing full-image dependencies. 3) The state-of-the-art performance. We conduct extensive experiments on popular semantic segmentation benchmarks including Cityscapes, ADE20K, and instance segmentation benchmark COCO. In particular, our CCNet achieves the mIoU score of 81.4 and 45.22 on Cityscapes test set and ADE20K validation set, respectively, which are the new state-of-the-art results. The source code is available at https://github.com/speedinghzl/CCNet.

/pdf/ccnet-criss-cross-attention-for-semantic-segmentation-2m8k1ptda1.pdf

CCNet: Criss-Cross Attention for Semantic Segmentation

Datasets drive vision progress, yet existing driving datasets are impoverished in terms of visual content and supported tasks to study multitask learning for autonomous driving. Researchers are usually constrained to study a small set of problems on one dataset, while real-world computer vision applications require performing tasks of various complexities. We construct BDD100K, the largest driving video dataset with 100K videos and 10 tasks to evaluate the exciting progress of image recognition algorithms on autonomous driving. The dataset possesses geographic, environmental, and weather diversity, which is useful for training models that are less likely to be surprised by new conditions. Based on this diverse dataset, we build a benchmark for heterogeneous multitask learning and study how to solve the tasks together. Our experiments show that special training strategies are needed for existing models to perform such heterogeneous tasks. BDD100K opens the door for future studies in this important venue.

/pdf/bdd100k-a-diverse-driving-dataset-for-heterogeneous-1v7dodhb6h.pdf

BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning

Incorporating multi-scale features in fully convolutional neural networks (FCNs) has been a key element to achieving state-of-the-art performance on semantic image segmentation. One common way to extract multi-scale features is to feed multiple resized input images to a shared deep network and then merge the resulting features for pixelwise classification. In this work, we propose an attention mechanism that learns to softly weight the multi-scale features at each pixel location. We adapt a state-of-the-art semantic image segmentation model, which we jointly train with multi-scale input images and the attention model. The proposed attention model not only outperforms averageand max-pooling, but allows us to diagnostically visualize the importance of features at different positions and scales. Moreover, we show that adding extra supervision to the output at each scale is essential to achieving excellent performance when merging multi-scale features. We demonstrate the effectiveness of our model with extensive experiments on three challenging datasets, including PASCAL-Person-Part, PASCAL VOC 2012 and a subset of MS-COCO 2014.

/pdf/attention-to-scale-scale-aware-semantic-image-segmentation-5fn97t6map.pdf

Attention to Scale: Scale-Aware Semantic Image Segmentation

With the objective to improve road safety, the automotive industry is moving toward more “intelligent” vehicles. One of the major challenges is to detect dangerous situations and react accordingly in order to avoid or mitigate accidents. This requires predicting the likely evolution of the current traffic situation, and assessing how dangerous that future situation might be. This paper is a survey of existing methods for motion prediction and risk assessment for intelligent vehicles. The proposed classification is based on the semantics used to define motion and risk. We point out the tradeoff between model completeness and real-time constraints, and the fact that the choice of a risk assessment method is influenced by the selected motion model.

/pdf/a-survey-on-motion-prediction-and-risk-assessment-for-39io9df2mm.pdf

A survey on motion prediction and risk assessment for intelligent vehicles

Detecting the road area and ego-lane ahead of a vehicle is central to modern driver assistance systems. While lane-detection on well-marked roads is already available in modern vehicles, finding the boundaries of unmarked or weakly marked roads and lanes as they appear in inner-city and rural environments remains an unsolved problem due to the high variability in scene layout and illumination conditions, amongst others. While recent years have witnessed great interest in this subject, to date no commonly agreed upon benchmark exists, rendering a fair comparison amongst methods difficult. In this paper, we introduce a novel open-access dataset and benchmark for road area and ego-lane detection. Our dataset comprises 600 annotated training and test images of high variability from the KITTI autonomous driving project, capturing a broad spectrum of urban road scenes. For evaluation, we propose to use the 2D Bird's Eye View (BEV) space as vehicle control usually happens in this 2D world, requiring detection results to be represented in this very same space. Furthermore, we propose a novel, behavior-based metric which judges the utility of the extracted ego-lane area for driver assistance applications by fitting a driving corridor to the road detection results in the BEV. We believe this to be important for a meaningful evaluation as pixel-level performance is of limited value for vehicle control. State-of-the-art road detection algorithms are used to demonstrate results using classical pixel-level metrics in perspective and BEV space as well as the novel behavior-based performance measure. All data and annotations are made publicly available on the KITTI online evaluation website in order to serve as a common benchmark for road terrain detection algorithms.

/pdf/a-new-performance-measure-and-evaluation-benchmark-for-road-318r1gf0xs.pdf

A new performance measure and evaluation benchmark for road detection algorithms

In order to enable the widespread use of robots in home and office environments, systems with natural interaction capabilities have to be developed. A prerequisite for natural interaction is the robot's ability to automatically recognize when and how long a person's attention is directed towards it for communication. As in open environments several persons can be present simultaneously, the detection of the communication partner is of particular importance. In this paper we present an attention system for a mobile robot which enables the robot to shift its attention to the person of interest and to maintain attention during interaction. Our approach is based on a method for multi-modal person tracking which uses a pan-tilt camera for face recognition, two microphones for sound source localization, and a laser range finder for leg detection. Shifting of attention is realized by turning the camera into the direction of the person which is currently speaking. From the orientation of the head it is decided whether the speaker addresses the robot. The performance of the proposed approach is demonstrated with an evaluation. In addition, qualitative results from the performance of the robot at the exhibition part of the ICVS'03 are provided.

Providing the basis for human-robot-interaction: a multi-modal attention system for a mobile robot

Multi-modal anchoring for human–robot interaction

In the recent past, service robots that are able to interact with humans in a natural way have become increasingly popular. A special kind of service robots that are designed for personal use at home are the so-called robot companions. They are expected to communicate with non-expert users in natural and intuitive way. For such natural interactions with humans the robot has to detect communication partners and focus its attention on them. Moreover, the companion has to be able to understand speech and gestures of a user and to carry out dialogs in order to get instructed, i.e., introduced to its environment. We address these problems by presenting the current state of our mobile robot BIRON, the Bielefeld Robot Companion.

BIRON - The Bielefeld Robot Companion

A major challenge for the realization of intelligent robots is to supply them with cognitive abilities in order to allow ordinary users to program them easily and intuitively. One approach to such programming is teaching work tasks by interactive demonstration. To make this effective and convenient for the user, the machine must be capable of establishing a common focus of attention and be able to use and integrate spoken instructions, visual perception, and non-verbal clues like gestural commands. We report progress in building a hybrid architecture that combines statistical methods, neural networks, and finite state machines into an integrated system for instructing grasping tasks by man-machine interaction. The system combines the GRAVIS-robot for visual attention and gestural instruction with an intelligent interface for speech recognition and linguistic interpretation, and a modality fusion module to allow multi-modal task-oriented man-machine communication with respect to dextrous robot manipulation of objects.

/pdf/multi-modal-human-machine-communication-for-instructing-4wcwjnq669.pdf

Jannik Fritsch

Papers

A new performance measure and evaluation benchmark for road detection algorithms

Providing the basis for human-robot-interaction: a multi-modal attention system for a mobile robot

Multi-modal anchoring for human–robot interaction

BIRON - The Bielefeld Robot Companion

Multi-modal human-machine communication for instructing robot grasping tasks