We present a simple but effective pixel-level self-supervised distillation framework friendly to dense prediction tasks. Our method, called Pixel-Wise Contrastive Distillation (PCD), distills knowledge by attracting the corresponding pixels from student’s and teacher’s output feature maps. PCD includes a novel design called SpatialAdaptor which "reshapes" a part of the teacher network while preserving the distribution of its output features. Our ablation experiments suggest that this reshaping behavior enables more informative pixel-to-pixel distillation. Moreover, we utilize a plug-in multi-head self-attention module that explicitly relates the pixels of student’s feature maps to enhance the effective receptive field, leading to a more competitive student. PCD outperforms previous self-supervised distillation methods on various dense prediction tasks. A backbone of ResNet-18-FPN distilled by PCD achieves 37.4 APbboxand 34.0 APmaskon COCO dataset using the detector of Mask R-CNN. We hope our study will inspire future research on how to pre-train a small model friendly to dense prediction tasks in a self-supervised fashion.

We propose a new convolution called Dynamic Region-Aware Convolution (DRConv), which can automatically assign multiple filters to corresponding spatial regions where features have similar representation. In this way, DR-Conv outperforms standard convolution in modeling semantic variations. Standard convolutional layer can increase the number of filers to extract more visual elements but results in high computational cost. More gracefully, our DR-Conv transfers the increasing channel-wise filters to spatial dimension with learnable instructor, which not only improve representation ability of convolution, but also maintains computational cost and the translation-invariance as standard convolution dose. DRConv is an effective and elegant method for handling complex and variable spatial information distribution. It can substitute standard convolution in any existing networks for its plug-and-play property, especially to power convolution layers in efficient networks. We evaluate DRConv on a wide range of models (MobileNet series, ShuffleNetV2, etc.) and tasks (Classification, Face Recognition, Detection and Segmentation). On ImageNet classification, DRConv-based ShuffleNetV2-0.5 achieves state-of-the-art performance of 67.1% at 46M× multiply-adds level with 6.3% relative improvement.

In this paper, we propose a novel meta learning approach for automatic channel pruning of very deep neural networks. We first train a PruningNet, a kind of meta network, which is able to generate weight parameters for any pruned structure given the target network. We use a simple stochastic structure sampling method for training the PruningNet. Then, we apply an evolutionary procedure to search for good-performing pruned networks. The search is highly efficient because the weights are directly generated by the trained PruningNet and we do not need any finetuning at search time. With a single PruningNet trained for the target network, we can search for various Pruned Networks under different constraints with little human participation. Compared to the state-of-the-art pruning methods, we have demonstrated superior performances on MobileNet V1/V2 and ResNet. Codes are available on https://github.com/liuzechun/MetaPruning.

Most visual based obstacle detection methods usually detect obstacles on the whole image, which are influenced by complex background and various illuminations. This paper proposes a novel and simple obstacle detection method DLP based on analyzing Distortion of Laser Pattern. To reduce computation, a laser pattern with two cross lines is used to project onto the front ground. Firstly, a segmentation algorithm based on color and edge is proposed to extract laser pixels. Then, skeletons of laser pattern are quickly extracted using table scanning strategy to further reduce noise. Secondly, seed-filling is used to extract the region of skeletons, and then laser lines are detected by Hough and merged by region based rule. Finally, according to the distortion of laser pattern projected on different obstacles, we propose a hierarchical discriminant model, which can classify the front scene as free or having obstacles. And the obstacle can be further classified as raised or sunken obstacle, and ascending or descending stairs. Experimental results show the robustness and effectiveness of our DLP method in different scenes, including indoor and outdoor, day and night with 10f/s.