def parse_event(event_path, cfg): event = load_event(event_path, cfg) event = torch.from_numpy(event) # 考虑基于密度的去噪 denoise_events = getattr(cfg, 'denoise_events', False) denoise_bins = getattr(cfg, 'denoise_bins', 10) denoise_timeslice = getattr(cfg, 'denoise_timeslice', 5000) denoise_patch = getattr(cfg, 'denoise_patch', 2) denoise_thres = getattr(cfg, 'denoise_thres', 0.5) denoise_density = getattr(cfg, 'denoise_density', False) denoise_hot = getattr(cfg, 'denoise_hot', False) denoise_time = getattr(cfg, 'denoise_time', False) denoise_neglect_polarity = getattr(cfg, 'denoise_neglect_polarity', True) reshape = getattr(cfg, 'reshape', False) if reshape: reshape_method = getattr(cfg, 'reshape_method', 'no_sample') if reshape_method == 'no_sample': event = reshape_event_no_sample(event, SENSOR_H, SENSOR_W, IMAGE_H, IMAGE_W) elif reshape_method == 'sample': event = reshape_event_with_sample(event, SENSOR_H, SENSOR_W, IMAGE_H, IMAGE_W) elif reshape_method == 'unique': event = reshape_event_unique(event, SENSOR_H, SENSOR_W, IMAGE_H, IMAGE_W) # Account for slicing slice_events = getattr(cfg, 'slice_events', False) if slice_events: event = slice_event(event, cfg) return event这个函数什么意思