I have been trying to train a single-layer network on N-MNIST.
Got a code snippet from the Tonic dataset library:
dataset = tonic.datasets.NMNIST(save_to='./data', train=False) dataloader = torch.utils.data.DataLoader(dataset, shuffle=True, num_workers=4, pin_memory=True) events, target = next(iter(dataloader)) print(events.shape) print(target.shape)
I get output as:
torch.Size([1, 4517, 4]) torch.Size()
So essentially for a single image, I got 4517 events each with 4 fields: x (pixel), y(pixel), time_stamp(in us), polarity.
Did anyone try training any kind of network with such a dataset?