Scidata.CIFAR100.labels_info
You're seeing just the function
labels_info
, go back to Scidata.CIFAR100 module for more information.
Shows descriptions of coarse and fine labels of the dataset.
Label values returned by download/1
correspond to indices in the lists
returned here.
Examples
iex> {_, labels} = Scidata.CIFAR100.download()
{{<<255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 231, 176, 237, 255, 255, 255, 255, 255, 252, 242, 229, 195,
212, 182, 255, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254, 254,
254, 254, 254, ...>>, {:u, 8}, {50000, 3, 32, 32}},
{<<11, 19, 15, 29, 4, 0, 14, 11, 1, 1, 5, 86, 18, 90, 3, 28, 10, 23, 11, 31, 5,
39, 17, 96, 2, 82, 9, 17, 10, 71, 5, 39, 18, 8, 8, 97, 16, 80, 10, 71, 16,
74, 17, 59, 2, 70, 5, ...>>, {:u, 8}, {50000, 2}}}
iex> {coarse_class_names, fine_class_names} = Scidata.CIFAR100.labels_info()
{["aquatic_mammals", "fish", "flowers", "food_containers",
"fruit_and_vegetables", "household_electrical_devices", "household_furniture",
"insects", "large_carnivores", "large_man-made_outdoor_things",
"large_natural_outdoor_scenes", "large_omnivores_and_herbivores",
"medium_mammals", "non-insect_invertebrates", "people", "reptiles",
"small_mammals", "trees", "vehicles_1", "vehicles_2"],
["apple", "aquarium_fish", "baby", "bear", "beaver", "bed", "bee", "beetle",
"bicycle", "bottle", "bowl", "boy", "bridge", "bus", "butterfly", "camel",
"can", "castle", "caterpillar", "cattle", "chair", "chimpanzee", "clock",
"cloud", "cockroach", "couch", "crab", "crocodile", "cup", "dinosaur",
"dolphin", "elephant", "flatfish", "forest", "fox", "girl", "hamster",
"house", "kangaroo", "keyboard", "lamp", "lawn_mower", "leopard", "lion",
"lizard", "lobster", "man", "maple_tree", ...]}
iex> {labels_bin, labels_type, labels_shape} = labels
{<<11, 19, 15, 29, 4, 0, 14, 11, 1, 1, 5, 86, 18, 90, 3, 28, 10, 23, 11, 31, 5,
39, 17, 96, 2, 82, 9, 17, 10, 71, 5, 39, 18, 8, 8, 97, 16, 80, 10, 71, 16,
74, 17, 59, 2, 70, 5, 87, 17, ...>>, {:u, 8}, {50000, 2}}
iex> labels_tensor = labels_bin |> Nx.from_binary(labels_type) |> Nx.reshape(labels_shape)
#Nx.Tensor<
u8[50000][2]
[
[11, 19],
[15, 29],
[4, 0],
[14, 11],
[1, 1],
...
]
>
iex> coarse_labels = labels_tensor |> Nx.slice([0,0], [50000, 1]) |> Nx.reshape({50000}) |> Nx.to_flat_list() |> Enum.map(fn label_index -> Enum.at(coarse, label_index) end)
["large_omnivores_and_herbivores", "reptiles", "fruit_and_vegetables", "people",
"fish", "household_electrical_devices", "vehicles_1", "food_containers",
"large_natural_outdoor_scenes", "large_omnivores_and_herbivores", ...]
iex> fine_labels = labels_tensor |> Nx.slice([0,1], [50000, 1]) |> Nx.reshape({50000}) |> Nx.to_flat_list |> Enum.map(fn label_index -> Enum.at(fine, label_index) end)
["cattle", "dinosaur", "apple", "boy", "aquarium_fish", "telephone", "train",
"cup", "cloud", "elephant", "keyboard", "willow_tree", "sunflower", "castle", ...]
iex> Enum.zip(coarse_labels, fine_labels)
[
{"large_omnivores_and_herbivores", "cattle"},
{"reptiles", "dinosaur"},
{"fruit_and_vegetables", "apple"},
{"people", "boy"},
{"fish", "aquarium_fish"},
{"household_electrical_devices", "telephone"},
{"vehicles_1", "train"},
...
]