Search

We propose a passenger flow detection method for dense areas of subway stations to address the current situation that existing pedestrian detection models cannot meet the real-time performance requirements in subway applications and lack validation in multiple subway scenarios. First, we designed the MPDNet model, which uses PVT-small to extract features and an improved feature pyramid network (FPN) for upsampling using the adaptively spatial feature fusion (ASFF) algorithm to retain more local information in the output of the FPN. Second, to better evaluate the performance of models in the metro, we collected subway surveillance video data and proposed the MetroStation dataset. Finally, we trained and evaluated the performance of the MPDNet model on the MetroStation dataset. We compare our method with several common object detection models on the MetroStation dataset, using mAP and frames per second (FPS) to verify its accuracy. The experiments on the MetroStation dataset demonstrated that the MPDNet performed well and satisfied inference speed requirements in metro passenger flow detection.

Designing optimal reward functions has been desired but extremely difficult in reinforcement learning (RL). When it comes to modern complex tasks, sophisticated reward functions are widely used to simplify policy learning yet even a tiny adjustment on them is expensive to evaluate due to the drastically increasing cost of training. To this end, we propose a hindsight reward tweaking approach by designing a novel paradigm for deep reinforcement learning to model the influences of reward functions within a near-optimal space. We simply extend the input observation with a condition vector linearly correlated with the effective environment reward parameters and train the model in a conventional manner except for randomizing reward configurations, obtaining a hyper-policy whose characteristics are sensitively regulated over the condition space. We demonstrate the feasibility of this approach and study one of its potential application in policy performance boosting with multiple MuJoCo tasks.