search

tub-rip / event_based_optical_flow

The official implementation of "Secrets of Event-based Optical Flow" (ECCV2022 Oral and IEEE T-PAMI 2024)
GNU General Public License v3.0
138 stars 13 forks source link

Pretrained model and checkpoint file #4

Closed Sean0120 closed 1 year ago

Sean0120 commented 1 year ago

Hi, Thanks for your amazing work. Can you provide some pretrained models and checkpoint files and the config files for DSEC dataset?

shiba24 commented 1 year ago

Pretrained model -- sure. DSEC -- ok, might take some time but I can share!

shiba24 commented 1 year ago

Hey @Sean0120 , this is the pretrained MVSEC model: https://drive.google.com/file/d/13m-waAt5X0C7f0JLBwb6KAApYxgXoA2J/view?usp=sharing

I haven't tested if this is the exact one of the table 4 in the paper (although I think this is), so I will check it when I have time. After checking I will update readme.

Can you check if you can load and run inference?

DachunKai commented 1 year ago

@shiba24 Hey, shiba. Can you give some instructions about how to use this pretrained model to generate warped events or estimated flow such as Fig. 1. in the paper.

shiba24 commented 1 year ago

It's becoming a stale issue, I will close this, and update the readme with the link when I double check. Thanks!

shiba24 commented 1 year ago

@DachunKai I can provide the code for the inference using the model. But give me some time to clean up the code, so it may take some time. Btw, the Fig.1 in the paper is not output of DNNs, but from just model-based (optimization-based) method.