88: Shape of a tensor | TensorFlow | Tutorial

The video discusses intuition for tensor shapes for input to conv1d, conv2d and conv3d.

# --------------------
# Part - I: Intuition
# --------------------
00:00 - Visual intuition of 1d, 2d, and 3d inputs for convolution
12:40 - data_format: NWC, NHWC, NDHWC
14:10 - Create tensor of shape: (3,)
14:44 - Create tensor of shape: (3,1)
15:33 - Create tensor of shape: (1,3)
16:00 - Create tensor of shape: (3,3)
17:00 - Create tensor of shape: (4,3,3): NWC
18:08 - Slice tensor: NWC
19:31 - Create tensor of shape: (2,3)
20:17 - Create tensor of shape: (2,3,1)
20:37 - Create tensor of shape: (2,3,3)
21:29 - Create tensor of shape: (4,2,3,3): NHWC
22:15 - Slice tensor: NHWC
23:26 - Create tensor of shape: (1,2,3,3)
25:08 - Create tensor of shape: (1,4,2,3,3)
26:17 - Create tensor of shape: (5,4,2,3,3)
# -----------------------------
# Part - II: Image based tensor
# -----------------------------
30:30 - Slice and visualize image channels
38:50 - Reconstruct RGB image from individual channel tensors
# -----------------------------
# Part - III: Video based tensor
# -----------------------------
40:00 - 1 second video
40:42 - import cv2, functions
42:47 - Slice and visualize video to frames to channels
46:40 - Stack videos into a batch: (5,10,224,224,3)
49:44 - Ending notes

# ----------------
# TensorFlow Guide
# ----------------

Load video:

[wsasonorejo]

thank, very clear, very important for strong foundation in deep learning

[juliaclaira]

This is incredibly helpful. Thank you!