Datasets:

gvecchio

/

MatSynth

like 40

To load the dataset, you can use the HF APIs like in the example below. We will update the download script for extracting the raw materials.

ds = load_dataset(
    "gvecchio/MatSynth", 
    split = "test", 
    streaming = True,
)

Thank you for your response. May I have access to the training data? I was only able to download the test data.

Train data is available now and accessible in the same way!

Hello, I attempted to retrieve the image but received "Image(decode=True, id=None)".
Could you please advise if there is an issue with my code?

# load the dataset in streaming mode
ds = load_dataset(
    "gvecchio/MatSynth", 
    streaming = True,
)

# remove unnecessary columns to reduce downloaded data
ds = ds.remove_columns(["diffuse", "specular", "displacement", "opacity", "blend_mask"])
# keep only specified columns
ds = ds.select_columns(["basecolor", "normal", "roughness", "metallic"])

print(ds['train'].features['basecolor'])

Thank you!

Hello, print(ds['train'].features['basecolor']) is only printing you the type of the basecolor feature.
if you want to iterate over the trainset, you can do something like:

for x in ds["train"]:
    print(x["basecolor"])

Since you're loading the entire dataset the use ["train"] to select the train split. Alternatively, you can add split = "train" to the load_dataset call (same for the test set).
Let me know if you have further issues downloading the data. Soon enough I will release a dataset class for an easy integration with torch training code.