Max Zhaoshuo Li 李赵硕
Max Zhaoshuo Li 李赵硕
Hi @ZirongChan Yes, your result indeed is missing the details. Can you provide the following? - The commit you are on - A screenshot of the training metrics (both losses...
Could you look into wandb logging? It will be beneficial to have them for debugging purposes. I wonder if something is wrong with the setup and some metrics have gone...
Hi @qq297110281 We scale the compute up by using more GPUs. We find it is important to scale the batch size up for better results (such as 16 for TnT)....
Hi @aeskandari68 Yes this is possible. If you look at the [dataloder](https://github.com/NVlabs/neuralangelo/blob/main/projects/neuralangelo/data.py#L37), you can specify the held-out set logic. Currently we simply use a subset of training set as the...
Hi @SWWdz The DTU scenes only observe the objects from one side and have a background that is almost all white, which results in ambiguous surroundings. We use object masks...
Hi @SWWdz You can refer to prior work's provided functionalities [here](https://github.com/fdarmon/NeuralWarp/blob/main/evaluation/mesh_filtering.py#L108)
Thank you very much @arknapit! Look forward to the fixed dataset.
Hi @arknapit, I wonder if you have any updates on this thread :smile: Thank you!
Thank you for your reply! @arknapit
Hi @arknapit I think the number of poses doesn't match the images. The log file has 644 poses but the number of images is 643.