Semantic outdoor scene understanding based on 3D LiDAR point clouds is a challenging task for autonomous driving due to the sparse and irregular data structure. This paper takes advantages of the uneven range distribution of different LiDAR laser beams to propose a range aware instance segmentation network, RangeSeg. RangeSeg uses a shared encoder backbone with two range dependent decoders. A heavy decoder only computes top of a range image where the far and small objects locate to improve small object detection accuracy, and a light decoder computes whole range image for low computational cost. The results are further clustered by the DBSCAN method with a resolution weighted distance function to get instance-level segmentation results. Experiments on the KITTI dataset show that RangeSeg outperforms the state-of-the-art semantic segmentation methods with enormous speedup and improves the instance-level segmentation performance on small and far objects. The whole RangeSeg pipeline meets the real time requirement on NVIDIA\textsuperscript{\textregistered} JETSON AGX Xavier with 19 frames per second in average.
Occlusion handling is one of the challenges of object detection and segmentation, and scene understanding. Because objects appear differently when they are occluded in varying degree, angle, and locations. Therefore, determining the existence of occlusion between objects and their order in a scene is a fundamental requirement for semantic understanding. Existing works mostly use deep learning based models to retrieve the order of the instances in an image or for occlusion detection. This requires labelled occluded data and it is time consuming. In this paper, we propose a simpler and faster method that can perform both operations without any training and only requires the modal segmentation masks. For occlusion detection, instead of scanning the two objects entirely, we only focus on the intersected area between their bounding boxes. Similarly, we use the segmentation mask inside the same area to recover the depth-ordering. When tested on COCOA dataset, our method achieves +8% and +5% more accuracy than the baselines in order recovery and occlusion detection respectively.
In this paper, we extend scene understanding to include that of human sketch. The result is a complete trilogy of scene representation from three diverse and complementary modalities -- sketch, photo, and text. Instead of learning a rigid three-way embedding and be done with it, we focus on learning a flexible joint embedding that fully supports the ``optionality" that this complementarity brings. Our embedding supports optionality on two axes: (i) optionality across modalities -- use any combination of modalities as query for downstream tasks like retrieval, (ii) optionality across tasks -- simultaneously utilising the embedding for either discriminative (e.g., retrieval) or generative tasks (e.g., captioning). This provides flexibility to end-users by exploiting the best of each modality, therefore serving the very purpose behind our proposal of a trilogy in the first place. First, a combination of information-bottleneck and conditional invertible neural networks disentangle the modality-specific component from modality-agnostic in sketch, photo, and text. Second, the modality-agnostic instances from sketch, photo, and text are synergised using a modified cross-attention. Once learned, we show our embedding can accommodate a multi-facet of scene-related tasks, including those enabled for the first time by the inclusion of sketch, all without any task-specific modifications. Project Page: \url{http://www.pinakinathc.me/scenetrilogy}
3D object detection from multiple image views is a fundamental and challenging task for visual scene understanding. Due to its low cost and high efficiency, multi-view 3D object detection has demonstrated promising application prospects. However, accurately detecting objects through perspective views in the 3D space is extremely difficult due to the lack of depth information. Recently, DETR3D introduces a novel 3D-2D query paradigm in aggregating multi-view images for 3D object detection and achieves state-of-the-art performance. In this paper, with intensive pilot experiments, we quantify the objects located at different regions and find that the "truncated instances" (i.e., at the border regions of each image) are the main bottleneck hindering the performance of DETR3D. Although it merges multiple features from two adjacent views in the overlapping regions, DETR3D still suffers from insufficient feature aggregation, thus missing the chance to fully boost the detection performance. In an effort to tackle the problem, we propose Graph-DETR3D to automatically aggregate multi-view imagery information through graph structure learning (GSL). It constructs a dynamic 3D graph between each object query and 2D feature maps to enhance the object representations, especially at the border regions. Besides, Graph-DETR3D benefits from a novel depth-invariant multi-scale training strategy, which maintains the visual depth consistency by simultaneously scaling the image size and the object depth. Extensive experiments on the nuScenes dataset demonstrate the effectiveness and efficiency of our Graph-DETR3D. Notably, our best model achieves 49.5 NDS on the nuScenes test leaderboard, achieving new state-of-the-art in comparison with various published image-view 3D object detectors.
Task specification is at the core of programming autonomous robots. A low-effort modality for task specification is critical for engagement of non-expert end-users and ultimate adoption of personalized robot agents. A widely studied approach to task specification is through goals, using either compact state vectors or goal images from the same robot scene. The former is hard to interpret for non-experts and necessitates detailed state estimation and scene understanding. The latter requires the generation of desired goal image, which often requires a human to complete the task, defeating the purpose of having autonomous robots. In this work, we explore alternate and more general forms of goal specification that are expected to be easier for humans to specify and use such as images obtained from the internet, hand sketches that provide a visual description of the desired task, or simple language descriptions. As a preliminary step towards this, we investigate the capabilities of large scale pre-trained models (foundation models) for zero-shot goal specification, and find promising results in a collection of simulated robot manipulation tasks and real-world datasets.
Dense captioning in 3D point clouds is an emerging vision-and-language task involving object-level 3D scene understanding. Apart from coarse semantic class prediction and bounding box regression as in traditional 3D object detection, 3D dense captioning aims at producing a further and finer instance-level label of natural language description on visual appearance and spatial relations for each scene object of interest. To detect and describe objects in a scene, following the spirit of neural machine translation, we propose a transformer-based encoder-decoder architecture, namely SpaCap3D, to transform objects into descriptions, where we especially investigate the relative spatiality of objects in 3D scenes and design a spatiality-guided encoder via a token-to-token spatial relation learning objective and an object-centric decoder for precise and spatiality-enhanced object caption generation. Evaluated on two benchmark datasets, ScanRefer and ReferIt3D, our proposed SpaCap3D outperforms the baseline method Scan2Cap by 4.94% and 9.61% in [email omitted], respectively. Our project page with source code and supplementary files is available at https://SpaCap3D.github.io/ .
Accurate scene understanding from multiple sensors mounted on cars is a key requirement for autonomous driving systems. Nowadays, this task is mainly performed through data-hungry deep learning techniques that need very large amounts of data to be trained. Due to the high cost of performing segmentation labeling, many synthetic datasets have been proposed. However, most of them miss the multi-sensor nature of the data, and do not capture the significant changes introduced by the variation of daytime and weather conditions. To fill these gaps, we introduce SELMA, a novel synthetic dataset for semantic segmentation that contains more than 30K unique waypoints acquired from 24 different sensors including RGB, depth, semantic cameras and LiDARs, in 27 different atmospheric and daytime conditions, for a total of more than 20M samples. SELMA is based on CARLA, an open-source simulator for generating synthetic data in autonomous driving scenarios, that we modified to increase the variability and the diversity in the scenes and class sets, and to align it with other benchmark datasets. As shown by the experimental evaluation, SELMA allows the efficient training of standard and multi-modal deep learning architectures, and achieves remarkable results on real-world data. SELMA is free and publicly available, thus supporting open science and research.
Given a question-image input, the Visual Commonsense Reasoning (VCR) model can predict an answer with the corresponding rationale, which requires inference ability from the real world. The VCR task, which calls for exploiting the multi-source information as well as learning different levels of understanding and extensive commonsense knowledge, is a cognition-level scene understanding task. The VCR task has aroused researchers' interest due to its wide range of applications, including visual question answering, automated vehicle systems, and clinical decision support. Previous approaches to solving the VCR task generally rely on pre-training or exploiting memory with long dependency relationship encoded models. However, these approaches suffer from a lack of generalizability and losing information in long sequences. In this paper, we propose a parallel attention-based cognitive VCR network PAVCR, which fuses visual-textual information efficiently and encodes semantic information in parallel to enable the model to capture rich information for cognition-level inference. Extensive experiments show that the proposed model yields significant improvements over existing methods on the benchmark VCR dataset. Moreover, the proposed model provides intuitive interpretation into visual commonsense reasoning.
Monocular depth estimation is vital for scene understanding and downstream tasks. We focus on the supervised setup, in which ground-truth depth is available only at training time. Based on knowledge about the high regularity of real 3D scenes, we propose a method that learns to selectively leverage information from coplanar pixels to improve the predicted depth. In particular, we introduce a piecewise planarity prior which states that for each pixel, there is a seed pixel which shares the same planar 3D surface with the former. Motivated by this prior, we design a network with two heads. The first head outputs pixel-level plane coefficients, while the second one outputs a dense offset vector field that identifies the positions of seed pixels. The plane coefficients of seed pixels are then used to predict depth at each position. The resulting prediction is adaptively fused with the initial prediction from the first head via a learned confidence to account for potential deviations from precise local planarity. The entire architecture is trained end-to-end thanks to the differentiability of the proposed modules and it learns to predict regular depth maps, with sharp edges at occlusion boundaries. An extensive evaluation of our method shows that we set the new state of the art in supervised monocular depth estimation, surpassing prior methods on NYU Depth-v2 and on the Garg split of KITTI. Our method delivers depth maps that yield plausible 3D reconstructions of the input scenes. Code is available at: https://github.com/SysCV/P3Depth
Monocular depth estimation is a fundamental task in computer vision and has drawn increasing attention. Recently, some methods reformulate it as a classification-regression task to boost the model performance, where continuous depth is estimated via a linear combination of predicted probability distributions and discrete bins. In this paper, we present a novel framework called BinsFormer, tailored for the classification-regression-based depth estimation. It mainly focuses on two crucial components in the specific task: 1) proper generation of adaptive bins and 2) sufficient interaction between probability distribution and bins predictions. To specify, we employ the Transformer decoder to generate bins, novelly viewing it as a direct set-to-set prediction problem. We further integrate a multi-scale decoder structure to achieve a comprehensive understanding of spatial geometry information and estimate depth maps in a coarse-to-fine manner. Moreover, an extra scene understanding query is proposed to improve the estimation accuracy, which turns out that models can implicitly learn useful information from an auxiliary environment classification task. Extensive experiments on the KITTI, NYU, and SUN RGB-D datasets demonstrate that BinsFormer surpasses state-of-the-art monocular depth estimation methods with prominent margins. Code and pretrained models will be made publicly available at \url{https://github.com/zhyever/Monocular-Depth-Estimation-Toolbox}.
Real-time holistic scene understanding would allow machines to interpret their surrounding in a much more detailed manner than is currently possible. While panoptic image segmentation methods have brought image segmentation closer to this goal, this information has to be described relative to the 3D environment for the machine to be able to utilise it effectively. In this paper, we investigate methods for sequentially reconstructing static environments from panoptic image segmentations in 3D. We specifically target real-time operation: the algorithm must process data strictly online and be able to run at relatively fast frame rates. Additionally, the method should be scalable for environments large enough for practical applications. By applying a simple but powerful data-association algorithm, we outperform earlier similar works when operating purely online. Our method is also capable of reaching frame-rates high enough for real-time applications and is scalable to larger environments as well. Source code and further demonstrations are released to the public at: \url{https://tutvision.github.io/Online-Panoptic-3D/}
Semantic scene reconstruction from point cloud is an essential and challenging task for 3D scene understanding. This task requires not only to recognize each instance in the scene, but also to recover their geometries based on the partial observed point cloud. Existing methods usually attempt to directly predict occupancy values of the complete object based on incomplete point cloud proposals from a detection-based backbone. However, this framework always fails to reconstruct high fidelity mesh due to the obstruction of various detected false positive object proposals and the ambiguity of incomplete point observations for learning occupancy values of complete objects. To circumvent the hurdle, we propose a Disentangled Instance Mesh Reconstruction (DIMR) framework for effective point scene understanding. A segmentation-based backbone is applied to reduce false positive object proposals, which further benefits our exploration on the relationship between recognition and reconstruction. Based on the accurate proposals, we leverage a mesh-aware latent code space to disentangle the processes of shape completion and mesh generation, relieving the ambiguity caused by the incomplete point observations. Furthermore, with access to the CAD model pool at test time, our model can also be used to improve the reconstruction quality by performing mesh retrieval without extra training. We thoroughly evaluate the reconstructed mesh quality with multiple metrics, and demonstrate the superiority of our method on the challenging ScanNet dataset. Code is available at \url{https://github.com/ashawkey/dimr}.
Grounded situation recognition is the task of predicting the main activity, entities playing certain roles within the activity, and bounding-box groundings of the entities in the given image. To effectively deal with this challenging task, we introduce a novel approach where the two processes for activity classification and entity estimation are interactive and complementary. To implement this idea, we propose Collaborative Glance-Gaze TransFormer (CoFormer) that consists of two modules: Glance transformer for activity classification and Gaze transformer for entity estimation. Glance transformer predicts the main activity with the help of Gaze transformer that analyzes entities and their relations, while Gaze transformer estimates the grounded entities by focusing only on the entities relevant to the activity predicted by Glance transformer. Our CoFormer achieves the state of the art in all evaluation metrics on the SWiG dataset. Training code and model weights are available at https://github.com/jhcho99/CoFormer.
Despite the recent progress in deep learning, most approaches still go for a silo-like solution, focusing on learning each task in isolation: training a separate neural network for each individual task. Many real-world problems, however, call for a multi-modal approach and, therefore, for multi-tasking models. Multi-task learning (MTL) aims to leverage useful information across tasks to improve the generalization capability of a model. This thesis is concerned with multi-task learning in the context of computer vision. First, we review existing approaches for MTL. Next, we propose several methods that tackle important aspects of multi-task learning. The proposed methods are evaluated on various benchmarks. The results show several advances in the state-of-the-art of multi-task learning. Finally, we discuss several possibilities for future work.
In this paper, we focus on the Audio-Visual Question Answering (AVQA) task, which aims to answer questions regarding different visual objects, sounds, and their associations in videos. The problem requires comprehensive multimodal understanding and spatio-temporal reasoning over audio-visual scenes. To benchmark this task and facilitate our study, we introduce a large-scale MUSIC-AVQA dataset, which contains more than 45K question-answer pairs covering 33 different question templates spanning over different modalities and question types. We develop several baselines and introduce a spatio-temporal grounded audio-visual network for the AVQA problem. Our results demonstrate that AVQA benefits from multisensory perception and our model outperforms recent A-, V-, and AVQA approaches. We believe that our built dataset has the potential to serve as testbed for evaluating and promoting progress in audio-visual scene understanding and spatio-temporal reasoning. Code and dataset: http://gewu-lab.github.io/MUSIC-AVQA/
Automating video-based data and machine learning pipelines poses several challenges including metadata generation for efficient storage and retrieval and isolation of key-frames for scene understanding tasks. In this work, we present two semi-supervised approaches that automate this process of manual frame sifting in video streams by automatically classifying scenes for content and filtering frames for fine-tuning scene understanding tasks. The first rule-based method starts from a pre-trained object detector and it assigns scene type, uncertainty and lighting categories to each frame based on probability distributions of foreground objects. Next, frames with the highest uncertainty and structural dissimilarity are isolated as key-frames. The second method relies on the simCLR model for frame encoding followed by label-spreading from 20% of frame samples to label the remaining frames for scene and lighting categories. Also, clustering the video frames in the encoded feature space further isolates key-frames at cluster boundaries. The proposed methods achieve 64-93% accuracy for automated scene categorization for outdoor image videos from public domain datasets of JAAD and KITTI. Also, less than 10% of all input frames can be filtered as key-frames that can then be sent for annotation and fine tuning of machine vision algorithms. Thus, the proposed framework can be scaled to additional video data streams for automated training of perception-driven systems with minimal training images.
Texts in scene images convey critical information for scene understanding and reasoning. The abilities of reading and reasoning matter for the model in the text-based visual question answering (TextVQA) process. However, current TextVQA models do not center on the text and suffer from several limitations. The model is easily dominated by language biases and optical character recognition (OCR) errors due to the absence of semantic guidance in the answer prediction process. In this paper, we propose a novel Semantics-Centered Network (SC-Net) that consists of an instance-level contrastive semantic prediction module (ICSP) and a semantics-centered transformer module (SCT). Equipped with the two modules, the semantics-centered model can resist the language biases and the accumulated errors from OCR. Extensive experiments on TextVQA and ST-VQA datasets show the effectiveness of our model. SC-Net surpasses previous works with a noticeable margin and is more reasonable for the TextVQA task.
Aiming for higher-level scene understanding, this work presents a neural network approach that takes a road-layout map in bird's-eye-view as input, and predicts a human-interpretable graph that represents the road's topological layout. Our approach elevates the understanding of road layouts from pixel level to the level of graphs. To achieve this goal, an image-graph-image auto-encoder is utilized. The network is designed to learn to regress the graph representation at its auto-encoder bottleneck. This learning is self-supervised by an image reconstruction loss, without needing any external manual annotations. We create a synthetic dataset containing common road layout patterns and use it for training of the auto-encoder in addition to the real-world Argoverse dataset. By using this additional synthetic dataset, which conceptually captures human knowledge of road layouts and makes this available to the network for training, we are able to stabilize and further improve the performance of topological road layout understanding on the real-world Argoverse dataset. The evaluation shows that our approach exhibits comparable performance to a strong fully-supervised baseline.
Promising performance has been achieved for visual perception on the point cloud. However, the current methods typically rely on labour-extensive annotations on the scene scans. In this paper, we explore how synthetic models alleviate the real scene annotation burden, i.e., taking the labelled 3D synthetic models as reference for supervision, the neural network aims to recognize specific categories of objects on a real scene scan (without scene annotation for supervision). The problem studies how to transfer knowledge from synthetic 3D models to real 3D scenes and is named Referring Transfer Learning (RTL). The main challenge is solving the model-to-scene (from a single model to the scene) and synthetic-to-real (from synthetic model to real scene's object) gap between the synthetic model and the real scene. To this end, we propose a simple yet effective framework to perform two alignment operations. First, physical data alignment aims to make the synthetic models cover the diversity of the scene's objects with data processing techniques. Then a novel \textbf{convex-hull regularized feature alignment} introduces learnable prototypes to project the point features of both synthetic models and real scenes to a unified feature space, which alleviates the domain gap. These operations ease the model-to-scene and synthetic-to-real difficulty for a network to recognize the target objects on a real unseen scene. Experiments show that our method achieves the average mAP of 46.08\% and 55.49\% on the ScanNet and S3DIS datasets by learning the synthetic models from the ModelNet dataset. Code will be publicly available.
This paper presents a new vision Transformer, named Iwin Transformer, which is specifically designed for human-object interaction (HOI) detection, a detailed scene understanding task involving a sequential process of human/object detection and interaction recognition. Iwin Transformer is a hierarchical Transformer which progressively performs token representation learning and token agglomeration within irregular windows. The irregular windows, achieved by augmenting regular grid locations with learned offsets, 1) eliminate redundancy in token representation learning, which leads to efficient human/object detection, and 2) enable the agglomerated tokens to align with humans/objects with different shapes, which facilitates the acquisition of highly-abstracted visual semantics for interaction recognition. The effectiveness and efficiency of Iwin Transformer are verified on the two standard HOI detection benchmark datasets, HICO-DET and V-COCO. Results show our method outperforms existing Transformers-based methods by large margins (3.7 mAP gain on HICO-DET and 2.0 mAP gain on V-COCO) with fewer training epochs ($0.5 \times$).