Closed XiaRho closed 2 years ago
Hi, Please look at the newly added instructions for testing pre-trained models in the readme.
Thank you very much. One more question, when I try to reproduce the results (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting. I have already set the train_on_event_labels to True in settings_DSEC.yaml. However, the final MIoU is about 51.5. This seems to be the results of just using event labels. Could you also provide the pre-trained weights (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting?
Thank you very much. One more question, when I try to reproduce the results (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting. I have already set the train_on_event_labels to True in settings_DSEC.yaml. However, the final MIoU is about 51.5. This seems to be the results of just using event labels. Could you also provide the pre-trained weights (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting?
Hi, what are the total trainable parameters that ESS uses? I am unable to generate to number.
Thank you very much. One more question, when I try to reproduce the results (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting. I have already set the train_on_event_labels to True in settings_DSEC.yaml. However, the final MIoU is about 51.5. This seems to be the results of just using event labels. Could you also provide the pre-trained weights (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting?
Hi, what are the total trainable parameters that ESS uses? I am unable to generate to number.
I only set the model_name to ess_supervised (line 30), train_on_event_labels to True (line 35) in the setting_DSEC.yaml, and add this code self.use_task_a = True to the setting.py. I get the E2VID_lightweight from https://github.com/uzh-rpg/rpg_e2vid.
Hi @XiaRho, how much GPU memory is needed to run the experiment? What type of machines are you using? Thanks!
Hi @XiaRho, how much GPU memory is needed to run the experiment? What type of machines are you using? Thanks!
About 8000MB memory is need. l use the Tesla V100-SXM2-16GB.
Hi @XiaRho, how much GPU memory is needed to run the experiment? What type of machines are you using? Thanks!
About 8000MB memory is need. l use the Tesla V100-SXM2-16GB.
Thank you for your information!
Thank you very much. One more question, when I try to reproduce the results (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting. I have already set the train_on_event_labels to True in settings_DSEC.yaml. However, the final MIoU is about 51.5. This seems to be the results of just using event labels. Could you also provide the pre-trained weights (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting?
Hi We added the links for the pre-trained weights obtained in the supervised settings to the Google form.
Thank you very much. One more question, when I try to reproduce the results (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting. I have already set the train_on_event_labels to True in settings_DSEC.yaml. However, the final MIoU is about 51.5. This seems to be the results of just using event labels. Could you also provide the pre-trained weights (events+frames MIoU=53.29) on DSEC-Semantic in the supervised setting?
Hi We added the links for the pre-trained weights obtained in the supervised settings to the Google form.
Thank you very much!
Hello, I would like to know how to set up supervised learning on ess using only events? I have set train_on_event_labels: True and model_name: 'ess_supervised', but I still use frame for training...Thanks a lot!
Hi, after downloading the pre-trained weights, how to test your models to generate the semantic segmentation results and get the MIoU index?