Current 3D segmentation methods heavily rely on large-scale point-cloud datasets, which are notoriously laborious to annotate. Few attempts have been made to circumvent the need for dense per-point annotations. In this work, we look at weakly-supervised 3D semantic instance segmentation. The key idea is to leverage 3D bounding box labels which are easier and faster to annotate. Indeed, we show that it is possible to train dense segmentation models using only bounding box labels. At the core of our method, \name{}, lies a deep model, inspired by classical Hough voting, that directly votes for bounding box parameters, and a clustering method specifically tailored to bounding box votes. This goes beyond commonly used center votes, which would not fully exploit the bounding box annotations. On ScanNet test, our weakly supervised model attains leading performance among other weakly supervised approaches (+18 mAP@50). Remarkably, it also achieves 97% of the mAP@50 score of current fully supervised models. To further illustrate the practicality of our work, we train Box2Mask on the recently released ARKitScenes dataset which is annotated with 3D bounding boxes only, and show, for the first time, compelling 3D instance segmentation masks.
Neural architecture search (NAS) and network pruning are widely studied efficient AI techniques, but not yet perfect. NAS performs exhaustive candidate architecture search, incurring tremendous search cost. Though (structured) pruning can simply shrink model dimension, it remains unclear how to decide the per-layer sparsity automatically and optimally. In this work, we revisit the problem of layer-width optimization and propose Pruning-as-Search (PaS), an end-to-end channel pruning method to search out desired sub-network automatically and efficiently. Specifically, we add a depth-wise binary convolution to learn pruning policies directly through gradient descent. By combining the structural reparameterization and PaS, we successfully searched out a new family of VGG-like and lightweight networks, which enable the flexibility of arbitrary width with respect to each layer instead of each stage. Experimental results show that our proposed architecture outperforms prior arts by around $1.0\%$ top-1 accuracy under similar inference speed on ImageNet-1000 classification task. Furthermore, we demonstrate the effectiveness of our width search on complex tasks including instance segmentation and image translation. Code and models are released.
Self-supervised learning for computer vision has achieved tremendous progress and improved many downstream vision tasks such as image classification, semantic segmentation, and object detection. Among these, generative self-supervised vision learning approaches such as MAE and BEiT show promising performance. However, their global masked reconstruction mechanism is computationally demanding. To address this issue, we propose local masked reconstruction (LoMaR), a simple yet effective approach that performs masked reconstruction within a small window of 7$\times$7 patches on a simple Transformer encoder, improving the trade-off between efficiency and accuracy compared to global masked reconstruction over the entire image. Extensive experiments show that LoMaR reaches 84.1% top-1 accuracy on ImageNet-1K classification, outperforming MAE by 0.5%. After finetuning the pretrained LoMaR on 384$\times$384 images, it can reach 85.4% top-1 accuracy, surpassing MAE by 0.6%. On MS COCO, LoMaR outperforms MAE by 0.5 $\text{AP}^\text{box}$ on object detection and 0.5 $\text{AP}^\text{mask}$ on instance segmentation. LoMaR is especially more computation-efficient on pretraining high-resolution images, e.g., it is 3.1$\times$ faster than MAE with 0.2% higher classification accuracy on pretraining 448$\times$448 images. This local masked reconstruction learning mechanism can be easily integrated into any other generative self-supervised learning approach. Our code is publicly available in https://github.com/junchen14/LoMaR.
This paper addresses incremental few-shot instance segmentation, where a few examples of new object classes arrive when access to training examples of old classes is not available anymore, and the goal is to perform well on both old and new classes. We make two contributions by extending the common Mask-RCNN framework in its second stage -- namely, we specify a new object class classifier based on the probit function and a new uncertainty-guided bounding-box predictor. The former leverages Bayesian learning to address a paucity of training examples of new classes. The latter learns not only to predict object bounding boxes but also to estimate the uncertainty of the prediction as guidance for bounding box refinement. We also specify two new loss functions in terms of the estimated object-class distribution and bounding-box uncertainty. Our contributions produce significant performance gains on the COCO dataset over the state of the art -- specifically, the gain of +6 on the new classes and +16 on the old classes in the AP instance segmentation metric. Furthermore, we are the first to evaluate the incremental few-shot setting on the more challenging LVIS dataset.
In quantum machine field, detecting two-dimensional (2D) materials in Silicon chips is one of the most critical problems. Instance segmentation can be considered as a potential approach to solve this problem. However, similar to other deep learning methods, the instance segmentation requires a large scale training dataset and high quality annotation in order to achieve a considerable performance. In practice, preparing the training dataset is a challenge since annotators have to deal with a large image, e.g 2K resolution, and extremely dense objects in this problem. In this work, we present a novel method to tackle the problem of missing annotation in instance segmentation in 2D quantum material identification. We propose a new mechanism for automatically detecting false negative objects and an attention based loss strategy to reduce the negative impact of these objects contributing to the overall loss function. We experiment on the 2D material detection datasets, and the experiments show our method outperforms previous works.
High-resolution dense prediction enables many appealing real-world applications, such as computational photography, autonomous driving, etc. However, the vast computational cost makes deploying state-of-the-art high-resolution dense prediction models on hardware devices difficult. This work presents EfficientViT, a new family of high-resolution vision models with novel multi-scale linear attention. Unlike prior high-resolution dense prediction models that rely on heavy softmax attention, hardware-inefficient large-kernel convolution, or complicated topology structure to obtain good performances, our multi-scale linear attention achieves the global receptive field and multi-scale learning (two desirable features for high-resolution dense prediction) with only lightweight and hardware-efficient operations. As such, EfficientViT delivers remarkable performance gains over previous state-of-the-art models with significant speedup on diverse hardware platforms, including mobile CPU, edge GPU, and cloud GPU. Without performance loss on Cityscapes, our EfficientViT provides up to 13.9$\times$ and 6.2$\times$ GPU latency reduction over SegFormer and SegNeXt, respectively. For super-resolution, EfficientViT delivers up to 6.4x speedup over Restormer while providing 0.11dB gain in PSNR. For Segment Anything, EfficientViT delivers 48.9x higher throughput on A100 GPU while achieving slightly better zero-shot instance segmentation performance on COCO.
In this paper, we tackle the problem of learning visual representations from unlabeled scene-centric data. Existing works have demonstrated the potential of utilizing the underlying complex structure within scene-centric data; still, they commonly rely on hand-crafted objectness priors or specialized pretext tasks to build a learning framework, which may harm generalizability. Instead, we propose contrastive learning from data-driven semantic slots, namely SlotCon, for joint semantic grouping and representation learning. The semantic grouping is performed by assigning pixels to a set of learnable prototypes, which can adapt to each sample by attentive pooling over the feature and form new slots. Based on the learned data-dependent slots, a contrastive objective is employed for representation learning, which enhances the discriminability of features, and conversely facilitates grouping semantically coherent pixels together. Compared with previous efforts, by simultaneously optimizing the two coupled objectives of semantic grouping and contrastive learning, our approach bypasses the disadvantages of hand-crafted priors and is able to learn object/group-level representations from scene-centric images. Experiments show our approach effectively decomposes complex scenes into semantic groups for feature learning and significantly benefits downstream tasks, including object detection, instance segmentation, and semantic segmentation. Code is available at: https://github.com/CVMI-Lab/SlotCon.
Masked image modeling (MIM) learns representations with remarkably good fine-tuning performances, overshadowing previous prevalent pre-training approaches such as image classification, instance contrastive learning, and image-text alignment. In this paper, we show that the inferior fine-tuning performance of these pre-training approaches can be significantly improved by a simple post-processing in the form of feature distillation (FD). The feature distillation converts the old representations to new representations that have a few desirable properties just like those representations produced by MIM. These properties, which we aggregately refer to as optimization friendliness, are identified and analyzed by a set of attention- and optimization-related diagnosis tools. With these properties, the new representations show strong fine-tuning performance. Specifically, the contrastive self-supervised learning methods are made as competitive in fine-tuning as the state-of-the-art masked image modeling (MIM) algorithms. The CLIP models' fine-tuning performance is also significantly improved, with a CLIP ViT-L model reaching 89.0% top-1 accuracy on ImageNet-1K classification. On the 3-billion-parameter SwinV2-G model, the fine-tuning accuracy is improved by +1.5 mIoU / +1.1 mAP to 61.4 mIoU / 64.2 mAP on ADE20K semantic segmentation and COCO object detection, respectively, creating new records on both benchmarks. More importantly, our work provides a way for the future research to focus more effort on the generality and scalability of the learnt representations without being pre-occupied with optimization friendliness since it can be enhanced rather easily. The code will be available at https://github.com/SwinTransformer/Feature-Distillation.
Masked image modeling, an emerging self-supervised pre-training method, has shown impressive success across numerous downstream vision tasks with Vision transformers. Its underlying idea is simple: a portion of the input image is masked out and then reconstructed via a pre-text task. However, the working principle behind MIM is not well explained, and previous studies insist that MIM primarily works for the Transformer family but is incompatible with CNNs. In this work, we observe that MIM essentially teaches the model to learn better middle-order interactions among patches for more generalized feature extraction. We then propose an Architecture-Agnostic Masked Image Modeling framework (A$^2$MIM), which is compatible with both Transformers and CNNs in a unified way. Extensive experiments on popular benchmarks show that A$^2$MIM learns better representations without explicit design and endows the backbone model with the stronger capability to transfer to various downstream tasks.
Existing instance segmentation methods have achieved impressive performance but still suffer from a common dilemma: redundant representations (e.g., multiple boxes, grids, and anchor points) are inferred for one instance, which leads to multiple duplicated predictions. Thus, mainstream methods usually rely on a hand-designed non-maximum suppression (NMS) post-processing step to select the optimal prediction result, which hinders end-to-end training. To address this issue, we propose a box-free and NMS-free end-to-end instance segmentation framework, termed UniInst, that yields only one unique representation for each instance. Specifically, we design an instance-aware one-to-one assignment scheme, namely Only Yield One Representation (OYOR), which dynamically assigns one unique representation to each instance according to the matching quality between predictions and ground truths. Then, a novel prediction re-ranking strategy is elegantly integrated into the framework to address the misalignment between the classification score and the mask quality, enabling the learned representation to be more discriminative. With these techniques, our UniInst, the first FCN-based box-free and NMS-free instance segmentation framework, achieves competitive performance, e.g., 39.0 mask AP using ResNet-50-FPN and 40.2 mask AP using ResNet-101-FPN, against mainstream methods on COCO test-dev 2017. Moreover, the proposed instance-aware method is robust to occlusion scenes, outperforming common baselines by remarkable mask AP on the heavily-occluded OCHuman benchmark. Code is available at https://github.com/b03505036/UniInst.
Collecting image annotations remains a significant burden when deploying CNN in a specific applicative context. This is especially the case when the annotation consists in binary masks covering object instances. Our work proposes to delineate instances in three steps, based on a semi-automatic approach: (1) the extreme points of an object (left-most, right-most, top, bottom pixels) are manually defined, thereby providing the object bounding-box, (2) a universal automatic segmentation tool like Deep Extreme Cut is used to turn the bounded object into a segmentation mask that matches the extreme points; and (3) the predicted mask is manually corrected. Various strategies are then investigated to balance the human manual annotation resources between bounding-box definition and mask correction, including when the correction of instance masks is prioritized based on their overlap with other instance bounding-boxes, or the outcome of an instance segmentation model trained on a partially annotated dataset. Our experimental study considers a teamsport player segmentation task, and measures how the accuracy of the Panoptic-Deeplab instance segmentation model depends on the human annotation resources allocation strategy. It reveals that the sole definition of extreme points results in a model accuracy that would require up to 10 times more resources if the masks were defined through fully manual delineation of instances. When targeting higher accuracies, prioritizing the mask correction among the training set instances is also shown to save up to 80\% of correction annotation resources compared to a systematic frame by frame correction of instances, for a same trained instance segmentation model accuracy.
Thermal bridges are weak areas of building envelopes that conduct more heat to the outside than surrounding envelope areas. They lead to increased energy consumption and the formation of mold. With a neural network approach, we demonstrate a method of automatically detecting thermal bridges on building rooftops from panorama drone images of whole city districts. To train the neural network, we created a dataset including 917 images and 6895 annotations. The images in the dataset contain thermal information for detecting thermal bridges and a height map for rooftop recognition in addition to regular RGB information. Due to the small dataset, our approach currently only has an average recall of 9.4% @IoU:0.5-0.95 (14.4% for large objects). Nevertheless, our approach reliably detects structures only on rooftops and not on other parts of buildings, without any additional segmentation effort of building parts.
As an essential prerequisite for developing a medical intelligent assistant system, medical image segmentation has received extensive research and concentration from the neural network community. A series of UNet-like networks with encoder-decoder architecture has achieved extraordinary success, in which UNet2+ and UNet3+ redesign skip connections, respectively proposing dense skip connection and full-scale skip connection and dramatically improving compared with UNet in medical image segmentation. However, UNet2+ lacks sufficient information explored from the full scale, which will affect the learning of organs' location and boundary. Although UNet3+ can obtain the full-scale aggregation feature map, owing to the small number of neurons in the structure, it does not satisfy the segmentation of tiny objects when the number of samples is small. This paper proposes a novel network structure combining dense skip connections and full-scale skip connections, named UNet-sharp (UNet\#) for its shape similar to symbol \#. The proposed UNet\# can aggregate feature maps of different scales in the decoder sub-network and capture fine-grained details and coarse-grained semantics from the full scale, which benefits learning the exact location and accurately segmenting the boundary of organs or lesions. We perform deep supervision for model pruning to speed up testing and make it possible for the model to run on mobile devices; furthermore, designing two classification-guided modules to reduce false positives achieves more accurate segmentation results. Various experiments of semantic segmentation and instance segmentation on different modalities (EM, CT, MRI) and dimensions (2D, 3D) datasets, including the nuclei, brain tumor, liver, and lung, demonstrate that the proposed method outperforms state-of-the-art models.
This paper introduces a new matting task called human instance matting (HIM), which requires the pertinent model to automatically predict a precise alpha matte for each human instance. Straightforward combination of closely related techniques, namely, instance segmentation, soft segmentation and human/conventional matting, will easily fail in complex cases requiring disentangling mingled colors belonging to multiple instances along hairy and thin boundary structures. To tackle these technical challenges, we propose a human instance matting framework, called InstMatt, where a novel mutual guidance strategy working in tandem with a multi-instance refinement module is used, for delineating multi-instance relationship among humans with complex and overlapping boundaries if present. A new instance matting metric called instance matting quality (IMQ) is proposed, which addresses the absence of a unified and fair means of evaluation emphasizing both instance recognition and matting quality. Finally, we construct a HIM benchmark for evaluation, which comprises of both synthetic and natural benchmark images. In addition to thorough experimental results on complex cases with multiple and overlapping human instances each has intricate boundaries, preliminary results are presented on general instance matting. Code and benchmark are available in https://github.com/nowsyn/InstMatt.
Hierarchical clustering is an effective and efficient approach widely used for classical image segmentation methods. However, many existing methods using neural networks generate segmentation masks directly from per-pixel features, complicating the architecture design and degrading the interpretability. In this work, we propose a simpler, more interpretable architecture, called HCFormer. HCFormer accomplishes image segmentation by bottom-up hierarchical clustering and allows us to interpret, visualize, and evaluate the intermediate results as hierarchical clustering results. HCFormer can address semantic, instance, and panoptic segmentation with the same architecture because the pixel clustering is a common approach for various image segmentation tasks. In experiments, HCFormer achieves comparable or superior segmentation accuracy compared to baseline methods on semantic segmentation (55.5 mIoU on ADE20K), instance segmentation (47.1 AP on COCO), and panoptic segmentation (55.7 PQ on COCO).
Since the development of self-supervised visual representation learning from contrastive learning to masked image modeling (MIM), there is no significant difference in essence, that is, how to design proper pretext tasks for vision dictionary look-up. MIM recently dominates this line of research with state-of-the-art performance on vision Transformers (ViTs), where the core is to enhance the patch-level visual context capturing of the network via denoising auto-encoding mechanism. Rather than tailoring image tokenizers with extra training stages as in previous works, we unleash the great potential of contrastive learning on denoising auto-encoding and introduce a pure MIM method, ConMIM, to produce simple intra-image inter-patch contrastive constraints as the sole learning objectives for masked patch prediction. We further strengthen the denoising mechanism with asymmetric designs, including image perturbations and model progress rates, to improve the network pre-training. ConMIM-pretrained models with various scales achieve competitive results on downstream image classification, semantic segmentation, object detection, and instance segmentation tasks, e.g., on ImageNet-1K classification, we achieve 83.9% top-1 accuracy with ViT-Small and 85.3% with ViT-Base without extra data for pre-training.
Geometry diagram parsing plays a key role in geometry problem solving, wherein the primitive extraction and relation parsing remain challenging due to the complex layout and between-primitive relationship. In this paper, we propose a powerful diagram parser based on deep learning and graph reasoning. Specifically, a modified instance segmentation method is proposed to extract geometric primitives, and the graph neural network (GNN) is leveraged to realize relation parsing and primitive classification incorporating geometric features and prior knowledge. All the modules are integrated into an end-to-end model called PGDPNet to perform all the sub-tasks simultaneously. In addition, we build a new large-scale geometry diagram dataset named PGDP5K with primitive level annotations. Experiments on PGDP5K and an existing dataset IMP-Geometry3K show that our model outperforms state-of-the-art methods in four sub-tasks remarkably. Our code, dataset and appendix material are available at https://github.com/mingliangzhang2018/PGDP.
This work investigates a simple yet powerful dense prediction task adapter for Vision Transformer (ViT). Unlike recently advanced variants that incorporate vision-specific inductive biases into their architectures, the plain ViT suffers inferior performance on dense predictions due to weak prior assumptions. To address this issue, we propose the ViT-Adapter, which allows plain ViT to achieve comparable performance to vision-specific transformers. Specifically, the backbone in our framework is a plain ViT that can learn powerful representations from large-scale multi-modal data. When transferring to downstream tasks, a pre-training-free adapter is used to introduce the image-related inductive biases into the model, making it suitable for these tasks. We verify ViT-Adapter on multiple dense prediction tasks, including object detection, instance segmentation, and semantic segmentation. Notably, without using extra detection data, our ViT-Adapter-L yields state-of-the-art 60.9 box AP and 53.0 mask AP on COCO test-dev. We hope that the ViT-Adapter could serve as an alternative for vision-specific transformers and facilitate future research. The code and models will be released at https://github.com/czczup/ViT-Adapter.
As a rising task, panoptic segmentation is faced with challenges in both semantic segmentation and instance segmentation. However, in terms of speed and accuracy, existing LiDAR methods in the field are still limited. In this paper, we propose a fast and high-performance LiDAR-based framework, referred to as Panoptic-PHNet, with three attractive aspects: 1) We introduce a clustering pseudo heatmap as a new paradigm, which, followed by a center grouping module, yields instance centers for efficient clustering without object-level learning tasks. 2) A knn-transformer module is proposed to model the interaction among foreground points for accurate offset regression. 3) For backbone design, we fuse the fine-grained voxel features and the 2D Bird's Eye View (BEV) features with different receptive fields to utilize both detailed and global information. Extensive experiments on both SemanticKITTI dataset and nuScenes dataset show that our Panoptic-PHNet surpasses state-of-the-art methods by remarkable margins with a real-time speed. We achieve the 1st place on the public leaderboard of SemanticKITTI and leading performance on the recently released leaderboard of nuScenes.
In the emerging field of video coding for machines, video datasets with pristine video quality and high-quality annotations are required for a comprehensive evaluation. However, existing video datasets with detailed annotations are severely limited in size and video quality. Thus, current methods have to either evaluate their codecs on still images or on already compressed data. To mitigate this problem, we propose an evaluation method based on pseudo ground-truth data from the field of semantic segmentation to the evaluation of video coding for machines. Through extensive evaluation, this paper shows that the proposed ground-truth-agnostic evaluation method results in an acceptable absolute measurement error below 0.7 percentage points on the Bjontegaard Delta Rate compared to using the true ground truth for mid-range bitrates. We evaluate on the three tasks of semantic segmentation, instance segmentation, and object detection. Lastly, we utilize the ground-truth-agnostic method to measure the coding performances of the VVC compared against HEVC on the Cityscapes sequences. This reveals that the coding position has a significant influence on the task performance.