|
|
|
|
|
import os |
|
|
|
import h5py |
|
import numpy as np |
|
import PIL.Image as Image |
|
|
|
script_dir = os.path.dirname(__file__) |
|
datafile_path = os.path.join(script_dir, "../raw/LLD-logo.hdf5") |
|
|
|
with h5py.File(datafile_path, "r") as throwaway: |
|
samples_count: int = len(throwaway["data"]) |
|
|
|
|
|
def gen_samples( |
|
labels: list[str] = ["data", "meta_data/names"], datafile_path: str = datafile_path |
|
): |
|
|
|
|
|
with h5py.File(datafile_path, "r") as hdf5_file: |
|
count = len(hdf5_file["data"]) |
|
|
|
i = 0 |
|
while i < count: |
|
result = {} |
|
|
|
if "data" in labels: |
|
shape = hdf5_file["shapes"][i] |
|
images = hdf5_file["data"][i][:, : shape[1], : shape[2]] |
|
|
|
result["data"] = images.astype(np.uint8) |
|
|
|
for label in [l for l in labels if l != "data"]: |
|
result[label] = hdf5_file[label][i] |
|
|
|
yield result |
|
|
|
i += 1 |
|
|
|
|
|
if __name__ == "__main__": |
|
sample = next(gen_samples()) |
|
name = sample["meta_data/names"] |
|
images = sample["data"] |
|
|
|
print(name) |
|
|
|
image_pil = Image.fromarray(images[2]) |
|
image_pil.show() |
|
|
