No abstract

Construction sites are dangerous due to the complex interaction of workers with equipment, building materials, vehicles, etc. As a kind of protective gear, hardhats are crucial for the safety of people on construction sites. Therefore, it is necessary for administrators to identify the people that do not wear hardhats and send out alarms to them. As manual inspection is labor-intensive and expensive, it is ideal to handle this issue by a real-time automatic detector. As such, in this paper, we present an end-to-end convolutional neural network to solve the problem of detecting if workers are wearing hardhats. The proposed method focuses on localizing a person’s head and deciding whether they are wearing a hardhat. The MobileNet model is employed as the backbone network, which allows the detector to run in real time. A top-down module is leveraged to enhance the feature-extraction process. Finally, heads with and without hardhats are detected on multi-scale features using a residual-block-based prediction module. Experimental results on a dataset that we have established show that the proposed method could produce an average precision of 87.4%/89.4% at 62 frames per second for detecting people without/with a hardhat worn on the head.

No abstract

Path attribution methods are a popular tool to interpret a visual model's prediction on an input. They integrate model gradients for the input features over a path defined between the input and a reference, thereby satisfying certain desirable theoretical properties. However, their reliability hinges on the choice of the reference. Moreover, they do not exhibit weak dependence on the input, which leads to counter-intuitive feature attribution mapping. We show that path-based attribution can account for the weak dependence property by choosing the reference from the local distribution of the input. We devise a method to identify the local input distribution and propose a technique to stochastically integrate the model gradients over the paths defined by the references sampled from that distribution. Our local path integration (LPI) method is found to consistently outperform existing path attribution techniques when evaluated on deep visual models. Contributing to the ongoing search of reliable evaluation metrics for the interpretation methods, we also introduce DiffID metric that uses the relative difference between insertion and deletion games to alleviate the distribution shift problem faced by existing metrics. Our code is available at https://github.com/ypeiyu/LPI.

Despite their great advancement, current pedestrian detection methods focus on single static images, which fail to employ richer information available from the video sequences. Compared with still images, videos can offer temporal information of objects in the time dimension, thus providing the potential to obtain more robust detection performance. Here, a novel pedestrian detection method based on visible part detection and temporal calibration is proposed. Specifically, a part-aware module to predict the visible body part of each pedestrian instance, which enables us to obtain precise motion information of partially occluded pedestrians in a video sequence, is first developed. Then, the temporal coherence for each pedestrian instance based on the predicted motion information is constructed. After that, an adaptive temporal calibration method is introduced to effectively calibrate the final detection result. This method on two video pedestrian detection benchmarks, that is, Caltech-New and MOT17Det, is evaluated. Experimental results show that this method performs favourably against existing pedestrian detection approaches.