download NMNIST and decode

contributing.md

@@ -0,0 +1 @@
+Add guidlines here

evis/datasets/__init__.py

@@ -0,0 +1 @@
+from .nmnist import NMNIST

evis/datasets/nmnist.py

@@ -0,0 +1,117 @@
+import pathlib
+
+import numpy as np
+from torchvision.datasets.utils import download_and_extract_archive
+
+
+class NMNIST:
+    """
+    N-MNIST dataset introduced in `Converting Static Image Datasets to Spiking Neuromorphic Datasets Using Saccades <https://www.frontiersin.org/articles/10.3389/fnins.2015.00437/full>` paper.
+
+    Dataset `homepage <https://www.garrickorchard.com/datasets/n-mnist>`
+
+    Parameters
+    -----------
+    root: string or Pathlib.Path object.
+        Root directory where dataset will be downloaded and extracted.
+    split: string
+        Must be one of `train` or `test`
+    donwload: bool
+        Whether to download the dataset or not.
+
+    """
+
+    base_foler = "NMNIST"
+
+    urls = {
+        "train": "https://data.mendeley.com/public-files/datasets/468j46mzdv/files/39c25547-014b-4137-a934-9d29fa53c7a0/file_downloaded",
+        "test": "https://data.mendeley.com/public-files/datasets/468j46mzdv/files/05a4d654-7e03-4c15-bdfa-9bb2bcbea494/file_downloaded",
+    }
+
+    md5s = {
+        "train": "20959b8e626244a1b502305a9e6e2031",
+        "test": "69ca8762b2fe404d9b9bad1103e97832",
+    }
+
+    def __init__(self, root, split="train", download=False):
+        self.root = root
+        assert split in ("train", "test"), TypeError
+        self._split = split
+        self._base_folder = pathlib.Path(self.root) / self.base_foler
+
+        if download:
+            self.download()
+
+        if not self._check_exist():
+            raise RuntimeError(
+                "Dataset not found, you can use `download=True` to download it."
+            )
+
+    def download(self):
+        """
+        Downloads the dataset and extract using utility function from torchvision.
+        """
+        if not self._check_exist():
+
+            download_and_extract_archive(
+                url=self.urls[self._split],
+                download_root=self._base_folder,
+                filename=self._split + ".zip",
+                md5=self.md5s[self._split],
+            )
+
+    def decode_binary_to_npz(self, file_decode_path, npz_file_path):
+        """
+        Decoding single binary file to npz file.
+        """
+        # ref - https://drive.google.com/drive/folders/16PYo5Jo3VlFC6-Lvw4c2hB-EAEf_egTL
+        with open(file_decode_path, "rb") as bin_f:
+
+            data = np.uint32(np.fromfile(bin_f, dtype=np.uint8))
+            x = data[0::5]
+            y = data[1::5]
+
+            remaining_data = data[2::5]
+
+            p = (remaining_data & 128) >> 7  #
+            ts = ((remaining_data & 127) << 16) | (data[3::5] << 8) | (data[4::5])
+
+            event_data = {"x": x, "y": y, "p": p, "ts": ts}
+            np.savez(
+                npz_file_path,
+                ts=event_data["ts"],
+                x=event_data["x"],
+                y=event_data["y"],
+                p=event_data["p"],
+            )
+
+    def decode(self):
+        """
+        Decodes all the binary files in npz format.
+        """
+        source_dir = self._base_folder / f"{self._split.capitalize()}"
+        target_dir = self._base_folder / f"{self._split.capitalize()}_npz"
+
+        for class_dir_name in source_dir.iterdir():
+            target_class_dir = target_dir / class_dir_name.parts[-1]
+            target_class_dir.mkdir(parents=True, exist_ok=True)
+            for bin_file in class_dir_name.iterdir():
+                target_file = target_class_dir / (
+                    bin_file.parts[-1].split(".")[0] + ".npz"
+                )
+
+                self.decode_binary_to_npz(bin_file, target_file)
+                bin_file.unlink()  # removes the binary file as the data is now stored in npz format.
+        print("Decoded all the files. ")
+
+    def _check_exist(self):
+        return (self._base_folder / self._split.capitalize()).is_dir()
+
+    def __getitem__(self, index):
+        # TODO logic for indexing
+        pass
+
+    def __len__(self):
+        # TODO this depends on __getitem__ method.
+
+        pass