Aggregating information from features across different layers is an essential operation for dense prediction models. Despite its limited expressiveness, feature concatenation dominates the choice of aggregation operations. In this paper, we introduce Attentive Feature Aggregation (AFA) to fuse different network layers with more expressive non-linear operations. AFA exploits both spatial and channel attention to compute weighted average of the layer activations. Inspired by neural volume rendering, we extend AFA with Scale-Space Rendering (SSR) to perform late fusion of multi-scale predictions. AFA is applicable to a wide range of existing network designs. Our experiments show consistent and significant improvements on challenging semantic segmentation benchmarks, including Cityscapes, BDD100K, and Mapillary Vistas, at negligible computational and parameter overhead. In particular, AFA improves the performance of the Deep Layer Aggregation (DLA) model by nearly 6% mIoU on Cityscapes. Our experimental analyses show that AFA learns to progressively refine segmentation maps and to improve boundary details, leading to new state-of-the-art results on boundary detection benchmarks on BSDS500 and NYUDv2. Code and video resources are available at http://vis.xyz/pub/dla-afa.
We introduce the problem of weakly supervised Multi-Object Tracking and Segmentation, i.e. joint weakly supervised instance segmentation and multi-object tracking, in which we do not provide any kind of mask annotation. To address it, we design a novel synergistic training strategy by taking advantage of multi-task learning, i.e. classification and tracking tasks guide the training of the unsupervised instance segmentation. For that purpose, we extract weak foreground localization information, provided by Grad-CAM heatmaps, to generate a partial ground truth to learn from. Additionally, RGB image level information is employed to refine the mask prediction at the edges of the objects. We evaluate our method on KITTI MOTS, the most representative benchmark for this task, reducing the performance gap on the MOTSP metric between the fully supervised and weakly supervised approach to just 12% and 12.7% for cars and pedestrians, respectively.
Crop-based training strategies decouple training resolution from GPU memory consumption, allowing the use of large-capacity panoptic segmentation networks on multi-megapixel images. Using crops, however, can introduce a bias towards truncating or missing large objects. To address this, we propose a novel crop-aware bounding box regression loss (CABB loss), which promotes predictions to be consistent with the visible parts of the cropped objects, while not over-penalizing them for extending outside of the crop. We further introduce a novel data sampling and augmentation strategy which improves generalization across scales by counteracting the imbalanced distribution of object sizes. Combining these two contributions with a carefully designed, top-down panoptic segmentation architecture, we obtain new state-of-the-art results on the challenging Mapillary Vistas (MVD), Indian Driving and Cityscapes datasets, surpassing the previously best approach on MVD by +4.5% PQ and +5.2% mAP.
Pseudo-LiDAR-based methods for monocular 3D object detection have generated large attention in the community due to performance gains showed on the KITTI3D benchmark dataset, in particular on the commonly reported validation split. This generated a distorted impression about the superiority of Pseudo-LiDAR approaches against methods working with RGB-images only. Our first contribution consists in rectifying this view by analysing and showing experimentally that the validation results published by Pseudo-LiDAR-based methods are substantially biased. The source of the bias resides in an overlap between the KITTI3D object detection validation set and the training/validation sets used to train depth predictors feeding Pseudo-LiDAR-based methods. Surprisingly, the bias remains also after geographically removing the overlap, revealing the presence of a more structured contamination. This leaves the test set as the only reliable mean of comparison, where published Pseudo-LiDAR-based methods do not excel. Our second contribution brings Pseudo-LiDAR-based methods back up in the ranking with the introduction of a 3D confidence prediction module. Thanks to the proposed architectural changes, our modified Pseudo-LiDAR-based methods exhibit extraordinary gains on the test scores (up to +8% 3D AP).
In this work we propose five concrete steps to improve the performance of optical flow algorithms. We carefully reviewed recently introduced innovations and well-established techniques in deep learning-based flow methods including i) pyramidal feature representations, ii) flow-based consistency checks, iii) cost volume construction practices or iv) distillation, and present extensions or alternatives to inhibiting factors we identified therein. We also show how changing the way gradients propagate in modern flow networks can lead to surprising boosts in performance. Finally, we contribute a novel feature that adaptively guides the learning process towards improving on under-performing flow predictions. Our findings are conceptually simple and easy to implement, yet result in compelling improvements on relevant error measures that we demonstrate via exhaustive ablations on datasets like Flying Chairs2, Flying Things, Sintel and KITTI. We establish new state-of-the-art results on the challenging Sintel and Kitti 2015 test datasets, and even show the portability of our findings to different optical flow and depth from stereo approaches.
While expensive LiDAR and stereo camera rigs have enabled the development of successful 3D object detection methods, monocular RGB-only approaches still lag significantly behind. Our work advances the state of the art by introducing MoVi-3D, a novel, single-stage deep architecture for monocular 3D object detection. At its core, MoVi-3D leverages geometrical information to generate synthetic views from virtual cameras at both, training and test time, resulting in normalized object appearance with respect to distance. Our synthetically generated views facilitate the detection task as they cut down the variability in visual appearance associated to objects placed at different distances from the camera. As a consequence, the deep model is relieved from learning depth-specific representations and its complexity can be significantly reduced. In particular we show that our proposed concept of exploiting virtual cameras enables us to set new state-of-the-art results on the popular KITTI3D benchmark using just a lightweight, single-stage architecture.
In this work we contribute a novel pipeline to automatically generate training data, and to improve over state-of-the-art multi-object tracking and segmentation (MOTS) methods. Our proposed tracklet mining algorithm turns raw street-level videos into high-fidelity MOTS training data, is scalable and overcomes the need of expensive and time-consuming manual annotation approaches. We leverage state-of-the-art instance segmentation results in combination with optical flow obtained from models also trained on automatically harvested training data. Our second major contribution is MOTSNet - a deep learning, tracking-by-detection architecture for MOTS - deploying a novel mask-pooling layer for improved object association over time. Training MOTSNet with our automatically extracted data leads to significantly improved sMOTSA scores on the novel KITTI MOTS dataset (+1.9%/+7.5% on cars/pedestrians). Even without learning from a single, manually annotated MOTS training example we still improve over prior state-of-the-art, confirming the compelling properties of our pipeline. On the MOTSChallenge dataset we improve by +4.1%, further confirming the efficacy of our proposed MOTSNet.
In this paper we propose an approach for monocular 3D object detection from a single RGB image, which leverages a novel disentangling transformation for 2D and 3D detection losses and a novel, self-supervised confidence score for 3D bounding boxes. Our proposed loss disentanglement has the twofold advantage of simplifying the training dynamics in the presence of losses with complex interactions of parameters, and sidestepping the issue of balancing independent regression terms. Our solution overcomes these issues by isolating the contribution made by groups of parameters to a given loss, without changing its nature. We further apply loss disentanglement to another novel, signed Intersection-over-Union criterion-driven loss for improving 2D detection results. Besides our methodological innovations, we critically review the AP metric used in KITTI3D, which emerged as the most important dataset for comparing 3D detection results. We identify and resolve a flaw in the 11-point interpolated AP metric, affecting all previously published detection results and particularly biases the results of monocular 3D detection. We provide extensive experimental evaluations and ablation studies on the KITTI3D and nuScenes datasets, setting new state-of-the-art results on object category car by large margins.
In this work we introduce a novel, CNN-based architecture that can be trained end-to-end to deliver seamless scene segmentation results. Our goal is to predict consistent semantic segmentation and detection results by means of a panoptic output format, going beyond the simple combination of independently trained segmentation and detection models. The proposed architecture takes advantage of a novel segmentation head that seamlessly integrates multi-scale features generated by a Feature Pyramid Network with contextual information conveyed by a light-weight DeepLab-like module. As additional contribution we review the panoptic metric and propose an alternative that overcomes its limitations when evaluating non-instance categories. Our proposed network architecture yields state-of-the-art results on three challenging street-level datasets, i.e. Cityscapes, Indian Driving Dataset and Mapillary Vistas.