input images with standard size like 720p

[gilotch]

trafficstars

Thank you for your amazing work! I have an interest to use the ENet depth completion model. It is working very well with the KITTI database, but when I try to feed it with images of different size (640x360) instead of (1216x352) I am facing this error: RuntimeError: Given groups=1, weight of size [32, 4, 5, 5], expected input[1, 5, 360, 640] to have 4 channels, but got 5 channels instead. Is the input layer dedicated to this form factor? What is your recommendation to adapt to input image of size 360x640 or 720p?

[JUGGHM]

Thanks for your interest! The input data is down-sampled 5 times so both the height and width must be integer multiple of 2^5=32. So you could crop the width to 352 pixels or pad it to 384 pixels.

[gilotch]

Thank you for your answer! I did follow your advice and center crop height down to 352 pixels for both inputs (rgb and depth). As there was still some remaining issues with an error mentioning 'channels', I checked the mode of my color image input and I was indeed in "RGBA". After converting the color image from "RGBA" to "RGB", the code finally worked normally.

from PIL import Image

def crop_center(pil_img, crop_width, crop_height):
    img_width, img_height = pil_img.size
    return pil_img.crop(((img_width - crop_width) // 2,
                         (img_height - crop_height) // 2,
                         (img_width + crop_width) // 2,
                         (img_height + crop_height) // 2))

def resize_image(image_path):
    im = Image.open(image_path)
    im_new = crop_center(im, 32*int(im.size[0]/32), 32*int(im.size[1]/32))
    if (im_new.mode == 'RGBA'):
        im_new = im_new.convert('RGB')
    im_new.save(image_path, quality=95)