Real-time multi-model multi-task (MMMT) workloads, a new form of deep learning inference workloads, are emerging for applications areas like extended reality (XR) to support metaverse use cases. These workloads combine user interactivity with computationally complex machine learning (ML) activities. Compared to standard ML applications, these ML workloads present unique difficulties and constraints. Real-time MMMT workloads impose heterogeneity and concurrency requirements on future ML systems and devices, necessitating the development of new capabilities. This paper begins with a discussion of the various characteristics of these real-time MMMT ML workloads and presents an ontology for evaluating the performance of future ML hardware for XR systems. Next, we present XRBench, a collection of MMMT ML tasks, models, and usage scenarios that execute these models in three representative ways: cascaded, concurrent, and cascaded-concurrency for XR use cases. Finally, we emphasize the need for new metrics that capture the requirements properly. We hope that our work will stimulate research and lead to the development of a new generation of ML systems for XR use cases.
We tackle the task of NeRF inversion for style-based neural radiance fields, (e.g., StyleNeRF). In the task, we aim to learn an inversion function to project an input image to the latent space of a NeRF generator and then synthesize novel views of the original image based on the latent code. Compared with GAN inversion for 2D generative models, NeRF inversion not only needs to 1) preserve the identity of the input image, but also 2) ensure 3D consistency in generated novel views. This requires the latent code obtained from the single-view image to be invariant across multiple views. To address this new challenge, we propose a two-stage encoder for style-based NeRF inversion. In the first stage, we introduce a base encoder that converts the input image to a latent code. To ensure the latent code is view-invariant and is able to synthesize 3D consistent novel view images, we utilize identity contrastive learning to train the base encoder. Second, to better preserve the identity of the input image, we introduce a refining encoder to refine the latent code and add finer details to the output image. Importantly note that the novelty of this model lies in the design of its first-stage encoder which produces the closest latent code lying on the latent manifold and thus the refinement in the second stage would be close to the NeRF manifold. Through extensive experiments, we demonstrate that our proposed two-stage encoder qualitatively and quantitatively exhibits superiority over the existing encoders for inversion in both image reconstruction and novel-view rendering.
Real-time tracking of 3D hand pose in world space is a challenging problem and plays an important role in VR interaction. Existing work in this space are limited to either producing root-relative (versus world space) 3D pose or rely on multiple stages such as generating heatmaps and kinematic optimization to obtain 3D pose. Moreover, the typical VR scenario, which involves multi-view tracking from wide \ac{fov} cameras is seldom addressed by these methods. In this paper, we present a unified end-to-end differentiable framework for multi-view, multi-frame hand tracking that directly predicts 3D hand pose in world space. We demonstrate the benefits of end-to-end differentiabilty by extending our framework with downstream tasks such as jitter reduction and pinch prediction. To demonstrate the efficacy of our model, we further present a new large-scale egocentric hand pose dataset that consists of both real and synthetic data. Experiments show that our system trained on this dataset handles various challenging interactive motions, and has been successfully applied to real-time VR applications.
We introduce Token Merging (ToMe), a simple method to increase the throughput of existing ViT models without needing to train. ToMe gradually combines similar tokens in a transformer using a general and light-weight matching algorithm that is as fast as pruning while being more accurate. Off-the-shelf, ToMe can 2x the throughput of state-of-the-art ViT-L @ 512 and ViT-H @ 518 models on images and 2.2x the throughput of ViT-L on video with only a 0.2-0.3% accuracy drop in each case. ToMe can also easily be applied during training, improving in practice training speed up to 2x for MAE fine-tuning on video. Training with ToMe further minimizes accuracy drop, leading to 2x the throughput of ViT-B on audio for only a 0.4% mAP drop. Qualitatively, we find that ToMe merges object parts into one token, even over multiple frames of video. Overall, ToMe's accuracy and speed are competitive with state-of-the-art on images, video, and audio.
Open-vocabulary semantic segmentation aims to segment an image into semantic regions according to text descriptions, which may not have been seen during training. Recent two-stage methods first generate class-agnostic mask proposals and then leverage pre-trained vision-language models, e.g., CLIP, to classify masked regions. We identify the performance bottleneck of this paradigm to be the pre-trained CLIP model, since it does not perform well on masked images. To address this, we propose to finetune CLIP on a collection of masked image regions and their corresponding text descriptions. We collect training data by mining an existing image-caption dataset (e.g., COCO Captions), using CLIP to match masked image regions to nouns in the image captions. Compared with the more precise and manually annotated segmentation labels with fixed classes (e.g., COCO-Stuff), we find our noisy but diverse dataset can better retain CLIP's generalization ability. Along with finetuning the entire model, we utilize the "blank" areas in masked images using a method we dub mask prompt tuning. Experiments demonstrate mask prompt tuning brings significant improvement without modifying any weights of CLIP, and it can further improve a fully finetuned model. In particular, when trained on COCO and evaluated on ADE20K-150, our best model achieves 29.6% mIoU, which is +8.5% higher than the previous state-of-the-art. For the first time, open-vocabulary generalist models match the performance of supervised specialist models in 2017 without dataset-specific adaptations.
While transformers have begun to dominate many tasks in vision, applying them to large images is still computationally difficult. A large reason for this is that self-attention scales quadratically with the number of tokens, which in turn, scales quadratically with the image size. On larger images (e.g., 1080p), over 60% of the total computation in the network is spent solely on creating and applying attention matrices. We take a step toward solving this issue by introducing Hydra Attention, an extremely efficient attention operation for Vision Transformers (ViTs). Paradoxically, this efficiency comes from taking multi-head attention to its extreme: by using as many attention heads as there are features, Hydra Attention is computationally linear in both tokens and features with no hidden constants, making it significantly faster than standard self-attention in an off-the-shelf ViT-B/16 by a factor of the token count. Moreover, Hydra Attention retains high accuracy on ImageNet and, in some cases, actually improves it.
Traditional computer vision models are trained to predict a fixed set of predefined categories. Recently, natural language has been shown to be a broader and richer source of supervision that provides finer descriptions to visual concepts than supervised "gold" labels. Previous works, such as CLIP, use InfoNCE loss to train a model to predict the pairing between images and text captions. CLIP, however, is data hungry and requires more than 400M image-text pairs for training. The inefficiency can be partially attributed to the fact that the image-text pairs are noisy. To address this, we propose OTTER (Optimal TransporT distillation for Efficient zero-shot Recognition), which uses online entropic optimal transport to find a soft image-text match as labels for contrastive learning. Based on pretrained image and text encoders, models trained with OTTER achieve strong performance with only 3M image text pairs. Compared with InfoNCE loss, label smoothing, and knowledge distillation, OTTER consistently outperforms these baselines in zero shot evaluation on Google Open Images (19,958 classes) and multi-labeled ImageNet 10K (10032 classes) from Tencent ML-Images. Over 42 evaluations on 7 different dataset/architecture settings x 6 metrics, OTTER outperforms (32) or ties (2) all baselines in 34 of them.
Neural Architecture Search (NAS) has been widely adopted to design accurate and efficient image classification models. However, applying NAS to a new computer vision task still requires a huge amount of effort. This is because 1) previous NAS research has been over-prioritized on image classification while largely ignoring other tasks; 2) many NAS works focus on optimizing task-specific components that cannot be favorably transferred to other tasks; and 3) existing NAS methods are typically designed to be "proxyless" and require significant effort to be integrated with each new task's training pipelines. To tackle these challenges, we propose FBNetV5, a NAS framework that can search for neural architectures for a variety of vision tasks with much reduced computational cost and human effort. Specifically, we design 1) a search space that is simple yet inclusive and transferable; 2) a multitask search process that is disentangled with target tasks' training pipeline; and 3) an algorithm to simultaneously search for architectures for multiple tasks with a computational cost agnostic to the number of tasks. We evaluate the proposed FBNetV5 targeting three fundamental vision tasks -- image classification, object detection, and semantic segmentation. Models searched by FBNetV5 in a single run of search have outperformed the previous stateof-the-art in all the three tasks: image classification (e.g., +1.3% ImageNet top-1 accuracy under the same FLOPs as compared to FBNetV3), semantic segmentation (e.g., +1.8% higher ADE20K val. mIoU than SegFormer with 3.6x fewer FLOPs), and object detection (e.g., +1.1% COCO val. mAP with 1.2x fewer FLOPs as compared to YOLOX).
Traditional computer vision models are trained to predict a fixed set of predefined categories. Recently, natural language has been shown to be a broader and richer source of supervision that provides finer descriptions to visual concepts than supervised "gold" labels. Previous works, such as CLIP, use a simple pretraining task of predicting the pairings between images and text captions. CLIP, however, is data hungry and requires more than 400M image text pairs for training. We propose a data-efficient contrastive distillation method that uses soft labels to learn from noisy image-text pairs. Our model transfers knowledge from pretrained image and sentence encoders and achieves strong performance with only 3M image text pairs, 133x smaller than CLIP. Our method exceeds the previous SoTA of general zero-shot learning on ImageNet 21k+1k by 73% relatively with a ResNet50 image encoder and DeCLUTR text encoder. We also beat CLIP by 10.5% relatively on zero-shot evaluation on Google Open Images (19,958 classes).