A question about data pre-processing

[sherlie831]

Hi, thank you for your work. I encountered an error while executing det.py and located the following code snippet： `class DSEC(BaseDataset): num_classes = 1

def __init__(self, opt, mode):
    self.ROOT_DATASET_PATH = os.path.join(opt.root_dir, 'data/DSEC')
    pkl_filename = 'DSEC-GT.pkl'
    super(DSEC, self).__init__(opt, mode, self.ROOT_DATASET_PATH, pkl_filename)

def imagefile(self, v, i):
    return os.path.join(self.ROOT_DATASET_PATH, 'RGB-images', v, '{:0>6}.png'.format(i))

def eventfile(self, v, i):
    return os.path.join(self.ROOT_DATASET_PATH, 'Event-images', v, '{:0>6}.png'.format(i))

def eventfile_30ms(self, v, i):
    return os.path.join(self.ROOT_DATASET_PATH, 'Event-images_30ms', v, '{:0>6}.png'.format(i))
def eventfile_50ms(self, v, i):
    return os.path.join(self.ROOT_DATASET_PATH, 'Event-images_50ms', v, '{:0>6}.png'.format(i))`

This requires that there are pre-processed data dirs (RGB-images, Event-images, Event-images_30ms, Event-images_50ms) under the self.ROOT_DATASET_PATH, but I did not find the data format or the specific method of aggregating events into event images in the paper and the README.md.

Could you kindly supplement the data description or pre-processing code? Thank you very much if you reply!

[shaoyu-liu]

我也没找到怎么处理这个问题，请问您解决了吗？

[ZZY-Zhou]

Hello,

Thank you for your interest in our work.

The generated event frames we used can be downloaded here.

And please do not forget to download and rename the rgb_calib in DSEC-MOD.

Hope could help.

[sherlie831]

Hello,

Thank you for your interest in our work.

The generated event frames we used can be downloaded here.

And please do not forget to download and rename the rgb_calib in DSEC-MOD.

Hope could help.

Hi, Thanks for your reply, it really helped.

Additionally, I am keenly interested in metrics evaluation and visualization. Specifically, I am looking to calculate Video mAP (V.mAP). The videoAP() function in ACT.py requires the variable "tubename," which is obtained by combining the inference directory and the video name with '_tubes.pkl'. Could you kindly guide me in obtaining this inference result?

Moreover, I'm curious to know if the purpose of the function in src/vis/vis_det.py is to visualize the detection boxes on the images.

Thank you in advance for your assistance.