search

ashawkey / objaverse_filter

naive filter of objaverse
MIT License
120 stars 2 forks source link

How to download dataset using kiuisobj_v1_merged_80K.csv #2

Closed FisherYuuri closed 5 months ago

FisherYuuri commented 8 months ago

Hi, I am really comfused about how to use kiuisobj_v1_merged_80K.csv to download your training dataset. I would appreciate if you could send me a tutorial or download the script!

LIRIIS commented 6 months ago

same question

lclichen commented 5 months ago

I think that's a simple question, but I can share my script if anyone needs it.

import objaverse
import pandas as pd
# pip install objaverse pandas
import multiprocessing
kiui_uids = pd.read_csv("kiuisobj_v1_merged_80K.csv", header=None)
processes = multiprocessing.cpu_count()
uids = kiui_uids[1].values.tolist()
# make sure you have enough disk capacity
objaverse.load_objects(uids, download_processes=processes)
LiaoLW commented 1 month ago

I think that's a simple question, but I can share my script if anyone needs it.

import objaverse
import pandas as pd
# pip install objaverse pandas
import multiprocessing
kiui_uids = pd.read_csv("kiuisobj_v1_merged_80K.csv", header=None)
processes = multiprocessing.cpu_count()
uids = kiui_uids[1].values.tolist()
# make sure you have enough disk capacity
objaverse.load_objects(uids, download_processes=processes)

Thank you for your help, objaverse API only downloads glb files. But the implementation of dataset requires RGB and camera pose, is there any pre-processing required before utilize the dataset to train LGM?