[PDF] Focus on Low-Resolution Information: Multi-Granular Information-Lossless Model for Low-Resolution Human Pose Estimation | Semantic Scholar (2024)

[PDF] Semantic Reader

This paper proposes a network that maintains high-resolution representations through the whole process of human pose estimation and empirically demonstrates the effectiveness of the network through the superior pose estimation results over two benchmark datasets: the COCO keypoint detection dataset and the MPII Human Pose dataset.

This work provides simple and effective baseline methods for pose estimation that are helpful for inspiring and evaluating new ideas for the field and achieved on challenging benchmarks.

This work presents a two-stage pose Distillation for Whole-body Pose estimators, named DWPose, to improve their effectiveness and efficiency and releases a series of models with different sizes, from tiny to large, for satisfying various downstream tasks.

The Human-Art dataset is introduced and contains 50k high-quality images with over 123k person instances from 5 natural and 15 artificial scenarios, which are annotated with bounding boxes, keypoints, self-contact points, and text information for humans represented in both 2D and 3D.

A novel human pose estimation framework termed DistilPose, which bridges the gaps between heatmap-based and regression-based methods and maximizes the transfer of knowledge from the teacher model to the student model through Token-distilling Encoder (TDE) and Simulated Heatmaps.

This paper presents a novel end-to-end framework with Explicit box Detection for multi-person Pose estimation, called ED-Pose, where it unifies the contextual learning between human-level and keypoint-level information and surpasses heatmap-based Top-down methods under the same backbone.

An attention similarity knowledge distillation approach, which transfers attention maps obtained from a high resolution (HR) network as a teacher into an LRnetwork as a student to boost LR recognition performance, outperforming state-of-the-art results by simply transferring well-constructed attention maps.

A new CNN building block called SPD-Conv is proposed in place of each strided convolution layer and each pooling layer, and it is shown that this approach significantly outperforms state-of-the-art deep learning models, especially on tougher tasks with low-resolution images and small objects.

The proposed PETR method views pose estimation as a hierarchical set prediction problem and effectively removes the need for many hand-crafted modules like RoI cropping, NMS and grouping post-processing, and largely overcomes the feature misalignment difficulty in pose estimation and improves the performance considerably.

...

...