Video as input about monodepth HOT 3 CLOSED

Never mind, I figured out the issue. The model accepts normalized test image where the pixel range is 0 to 1. Dividing the read image by 255 solved the issue.

from monodepth.

Here is what I am doing.
I have replaced the following commands
dataloader = MonodepthDataloader(args.data_path, args.filenames_file, params, args.dataset, args.mode)
left = dataloader.left_image_batch
right = dataloader.right_image_batch
with a placeholder for left and right
left = tf.placeholder(tf.float32, [2,256,512,3])
right = tf.placeholder(tf.float32, [2, 256, 512])
the model was then created based on these placeholders
model = MonodepthModel(params, args.mode, left, right)

The left placeholer was initialized with the test image as follows
image = cv2.imread("data/challenge01.jpg")
left1 = tf.image.resize_images(image, [256, 512])
leftp = tf.stack([left1, tf.image.flip_left_right(left1)])
ll=sess.run(leftp)

And finally I used feed_dict to replace the placeholder left with an actual image as 4D tensor
disp = sess.run(model.disp_left_est[0], feed_dict={left:ll})

The code actually runs successfully but doesn't generate a valid disparity map. The output is more like edge detection. Can you please tell what am I missing?
Thank you.

from monodepth.

Hi,
can yo send me your changes to be able to accept a video as input
Thanks

from monodepth.