Search

Most supervised learning-based pose estimation methods for stacked scenes are trained on massive synthetic datasets. In most cases, the challenge is that the learned network on the training dataset is no longer optimal on the testing dataset. To address this problem, we propose a pose regression network PPR-Net++. It transforms each scene point into a point in the centroid space, followed by a clustering process and a voting process. In the training phase, a mapping function between the network's critical parameter (i.e., the bandwidth of the clustering algorithm) and the compactness of the centroid distributions is obtained. This function is used to adapt the bandwidth between centroid distributions of two different domains. In addition, to further improve the pose estimation accuracy, the network also predicts the confidence of each point, based on its visibility and pose error. Only the points with high confidence have the right to vote for the final object pose. In experiments, our method is trained on the IPA synthetic dataset and compared with the state-of-the-art algorithm. When tested with the public synthetic Sileane dataset, our method is better in all eight objects, where five of them are improved by more than 5% in average precision (AP). On IPA real dataset, our method outperforms a large margin by 20%. This lays a solid foundation for robot grasping in industrial scenarios.

A robust feature point detection (FPD) tool is critical for sketch-based engineering modelling. We propose an FPD algorithm for freehand strokes based on a deep learning approach, denoted as FPD-DL. First, a point-wise neural network is trained to learn local and global features from points' multi-scale context images to classify each point of a stroke into a line-point or a curve-point. Then, a segment merging procedure is designed to extract primitives from the identified segments. It explores the distribution patterns of neighbouring segments. Finally, the tangent points and lost corners are added among consecutive primitives. The algorithm robustness and accuracy are experimented on a dataset with 1500 strokes of 20 shapes, which are grouped into two categories: set-by-person and set-by-shape. In the set-by-person category, the AON (all-or-nothing) accuracy of FPD-DL is 97%, compared to 94.3% of the state-of-the-art algorithm. In the set-by-shape category, the robustness to stroke shape is tested, which is similar to the practical usage scenario. The AON accuracy of FPD-DL is 95%, without too much fluctuation. Thus, the new proposed approach is robust to both stroke shape and stroke style.

Accurate object 6D pose estimation is a core task for robot bin-picking applications, especially when objects are randomly stacked with heavy occlusion. To address this problem, this paper proposes a simple but novel Point-wise Pose Regression Network (PPR-Net). For each point in the point cloud, the network regresses a 6D pose of the object instance that the point belongs to. We argue that the regressed poses of points from the same object instance should be located closely in pose space. Thus, these points can be clustered into different instances and their corresponding objects’ 6D poses can be estimated simultaneously. In our experiments, PPR-Net outperforms the state-of-the-art approach by 15% - 41% in average precision when evaluated on the benchmark Sileane dataset. In addition, it also works well in real world robot bin-picking tasks.