search

TRI-ML / packnet-sfm

TRI-ML Monocular Depth Estimation Repository
https://tri-ml.github.io/packnet-sfm/
MIT License
1.23k stars 242 forks source link

How to train on my own dataset? #164

Open porwalnaman01 opened 3 years ago

porwalnaman01 commented 3 years ago

Hi, I want to train this model on my own dataset which only has png images. It would be very kind of you if you can help me here. Thanks in advance!

VitorGuizilini-TRI commented 3 years ago

We have an "ImageDataset" that supports image folders as input, can you please give that a try?

TPEmist commented 3 years ago

How can we start with this "ImageDataset"? is there any proper instruction?

porwalnaman01 commented 3 years ago

It would be very great if you can provide training instructions for a dataset which only has images

porwalnaman01 commented 3 years ago

@TPEmist were you able to train on your own dataset?

TPEmist commented 3 years ago

@porwalnaman01 not really, I'll give it a try again this weekend. how about you?

luda1013 commented 2 years ago

Hallo, i could not install horovod directly and that problem i could solve with help google colab. So i installed the horovod within colab and it works

And since then i try to work in colab. My Uni has their own dataset they want to train and i transfered the train.py in colab

Unbenannt2 Unbenannt1

Next step is i need to upload my images and call the function train(file)

My Question is exactly like the title of this post :) Has anyone succeded in uploading and training your own images?