neuro_dataset.py 文件源码

def make_td_images(td, num_spikes, step_factor=1):
    """Generate set of images from the Temporal Difference (td) events by reading a number of unique spikes
    td is read from a binary file (refer to eventvision.Readxxx functions)
    td: eventvision.Events
    num_spikes: number of unique spikes to accumulate before generating an image
    step_factor: proportional amount to shift before generating the next image.
        1 would result in no overlapping events between images
        0.6 would result in the next image overlapping with 40% of the previous image

    returns array of images
    """
    assert isinstance(td, ev.Events)
    assert isinstance(num_spikes, (int, long))
    assert num_spikes > 0
    assert step_factor > 0

    #with timer.Timer() as my_timer:
    event_offset = 0
    images = []
    while event_offset + num_spikes < td.data.size:
        image = np.zeros((td.height, td.width), dtype=np.uint8)
        unique_spike_count = 0
        index_ptr = event_offset
        while (unique_spike_count < num_spikes) & (index_ptr < td.data.size):
            event = td.data[index_ptr]
            y = event.y
            x = event.x
            if image[y, x] == 0:
                image[y, x] = 255
                unique_spike_count += 1

            index_ptr += 1

        #cv2.imshow('img', img)
        #cv2.waitKey(1)
        if unique_spike_count < num_spikes:
            break

        images.append(image)

        #offset next image
        total_spikes_traversed = index_ptr - event_offset
        event_offset += math.floor(total_spikes_traversed * step_factor) + 1
    #print 'Making images out of bin file took %s seconds' % my_timer.secs

    return images