Question for inference on MVSEC dataset

[Tianbo-Pan]

Thanks for your interesting and excellent work! I have used your provided code and pre-trained model ('ramnet_sim2real.pth.tar'), and inferred on the MVSEC dataset. However, as I do not know the correct image size during inference. (In test.py, 167-170 rows)

And I failed to load the MVSEC data with your provided dataloader/dataset_mvsec.py as some indices files are missed(e.g., 'valid_depth_indices.npy' used in 374th row of dataset_mvsec.py). So I used the class 'SynchronizedFramesEventsDataset' in dataloader/dataset.py for data loading.

And the test result has some difference with the results reported in the paper. Can you give me some advice? Thanks!

[anusha-devulapally]

Hi,
It would be really helpful if you could please share on how you ran the inference. I tried the inference same way as you did by changing the data loader to 'SynchronizedFramesEventsDataset' but my numbers are in 4.80, 6.37 and 7.74m for 10,20 and 30m respectively for outdoor night1.
And also, what rgb folder are you using in mvsec, is it visensor_left_sync?