Multi-Object Tracking (MOT) has rapidly progressed with the development of object detection and reidentification. However, motion modeling, which facilitates object association by forecasting short-term trajectories with past observations, has been relatively underexplored in recent years. Current motion models in MOT typically assume that the object motion is linear in a small time window and needs continuous observations, so these methods are sensitive to occlusions and non-linear motion and require high frame-rate videos. In this work, we show that a simple motion model can obtain state-of-theart tracking performance without other cues like appearance. We emphasize the role of "observation" when recovering tracks from being lost and reducing the error accumulated by linear motion models during the lost period. We thus name the proposed method as Observation-Centric SORT, OC-SORT for short. It remains simple, online, and real-time but improves robustness over occlusion and nonlinear motion. It achieves 63.2 and 62.1 HOTA on MOT17 and MOT20, respectively, surpassing all published methods. It also sets new states of the art on KITTI Pedestrian Tracking and DanceTrack where the object motion is highly non-linear. The code and model are available at https://github.com/noahcao/OC_SORT.

No abstract

We address the issue of domain gap when making use of synthetic data to train a scene-specific object detector and pose estimator. While previous works have shown that the constraints of learning a scene-specific model can be leveraged to create geometrically and photometrically consistent synthetic data, care must be taken to design synthetic content which is as close as possible to the real-world data distribution. In this work, we propose to solve domain gap through the use of appearance randomization to generate a wide range of synthetic objects to span the space of realistic images for training. An ablation study of our results is presented to delineate the individual contribution of different components in the randomization process. We evaluate our method on VIRAT, UA-DETRAC, EPFL-Car datasets, where we demonstrate that using scene specific domain randomized synthetic data is better than fine-tuning off-the-shelf models on limited real data.

No abstract

Figure 1: Left: Our proposed capture setup consists of multiple egocentric cameras from wearable glasses and stationary secondary cameras. This setup is flexible and mobile, allowing us to generate high-quality multi-human 3D annotations for diverse in-the-wild settings. Center: Multiple synchronized egocentric views while playing soccer. Right: Synchronized secondary views (cropped) from the stationary cameras. All cameras are spatiotemporally localized in the world coordinate.

A key assumption of top-down human pose estimation approaches is their expectation of having a single person present in the input bounding box. This often leads to failures in crowded scenes with occlusions. We propose a novel solution to overcome the limitations of this fundamental assumption. Our Multi-Hypothesis Pose Network (MHPNet) allows for predicting multiple 2D poses within a given bounding box. We introduce a Multi-Hypothesis Attention Block (MHAB) that can adaptively modulate channel-wise feature responses for each hypothesis and is parameter efficient. We demonstrate the efficacy of our approach by evaluating on COCO, CrowdPose and OCHuman datasets. Specifically, we achieve 70.0 AP on CrowdPose and 42.5 AP on OCHuman test sets, a significant improvement of 2.4 AP and 6.5 AP over prior art, respectively. When using ground truth bounding boxes for inference, MHPNet achieves an improvement of 0.7 AP on COCO, 0.9 AP on CrowdPose, and 9.1 AP on OCHuman validation sets compared to HRNet. Interestingly, when fewer, high confidence bounding boxes are used, HRNet's performance degrades (by 5 AP) on OCHuman, whereas MHPNet maintains a relatively stable performance (drop of 1 AP) for the same inputs.

No abstract

ImageCenter Map (Ours) OCHMR (Ours) ROMP Figure 1. To handle severe person-person occlusion, our proposed method OCHMR conditions the deep network on image spatial context using predicted body centermaps. OCHMR is trained using multi-person mesh interpenetration and depth ordering losses. In comparison to bottom-up ROMP [58], top-down OCHMR outputs pixel aligned mesh estimates for each individual under occlusion.