Hi @RodrigoGantier, and thanks for sharing this work.It is really a fantastic job.
According to your code and paper, a fixed number ( seq_len * self.num_evs) of events have been converted to voxel grid. As a result, the sequence of voxel grid is asynchronous, which means it is difficult to calculate the exact timestamp of a set of voxel grid consisting of 5 BINS as you have mentioned in your code. So, without timestamps, how can we match the events and images in the style of many to many? How can we find the training set composed of reconstructed image and ground truth? As the Fig.8. shown in the paper, the outcomes have been compared between m2o and m2m. And the performance of m2o is superior. It confuses me a lot about how to fulfill the experiments, especially about how to get the supervisory information between reconstructed image and ground truth.
By the way, I am a rookie of DL as well as EVENT camera. So, chances are that the text I have mentioned above is not as precise as yours. Forgive me please!
I am sincerely waiting for your reply! THX a lot.
Hi @RodrigoGantier, and thanks for sharing this work.It is really a fantastic job. According to your code and paper, a fixed number ( seq_len * self.num_evs) of events have been converted to voxel grid. As a result, the sequence of voxel grid is asynchronous, which means it is difficult to calculate the exact timestamp of a set of voxel grid consisting of 5 BINS as you have mentioned in your code. So, without timestamps, how can we match the events and images in the style of many to many? How can we find the training set composed of reconstructed image and ground truth? As the Fig.8. shown in the paper, the outcomes have been compared between m2o and m2m. And the performance of m2o is superior. It confuses me a lot about how to fulfill the experiments, especially about how to get the supervisory information between reconstructed image and ground truth. By the way, I am a rookie of DL as well as EVENT camera. So, chances are that the text I have mentioned above is not as precise as yours. Forgive me please! I am sincerely waiting for your reply! THX a lot.