Scribd
Upload
94 views18 pages

RAFT: Efficient Optical Flow Estimation

The document outlines a video analysis project focused on RAFT, a deep neural network designed for accurate and efficient optical flow estimation between video frames. It details the architecture of RAFT, including feature extraction, visual similarity, and iterative updates, which are crucial for estimating pixel-level motion. The project emphasizes the importance of optical flow in understanding scene dynamics and outlines next steps for implementation.

Uploaded by

Iheb Alimi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
94 views18 pages

RAFT: Efficient Optical Flow Estimation

The document outlines a video analysis project focused on RAFT, a deep neural network designed for accurate and efficient optical flow estimation between video frames. It details the architecture of RAFT, including feature extraction, visual similarity, and iterative updates, which are crucial for estimating pixel-level motion. The project emphasizes the importance of optical flow in understanding scene dynamics and outlines next steps for implementation.

Uploaded by

Iheb Alimi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

VIDEO ANALYSIS PROJECT NOVEMBER 14, 2025

RAFT
Accurate & Efficient Optical Flow
Estimation
Agenda 1 Introduction

2 Optical Flow

3 RAFT Architecture

4 Training & Implementation


What is Optical Flow and Why It Matters

Estimates pixel-level motion between video frames


Captures dynamics in videos and images
Key for understanding scene movement

Figure 1. Optical Flow Estimation for a sequence of images. Frame 1, Frame 2, Computed Optical Flow between frame 1 and frame 2.
Architecture

RAFT is a Deep Neural Network that is


able to estimate the Dense Optical
Flow given a pair of sequential images
I₁ and I₂

LET'S BEGIN!

BACK TO AGENDA PAGE


An estimation of Dense Optical Flow assigns each pixel a 2D flow vector
describing it’s horizontal and vertical displacement over a time interval.

The brightness consistency assumption implies that a pixel at (x,y,t) will


have the same intensity at (x+dx, y+dy, t+dy). Therefore: I(x, y, t) = I(x+dx,
y+dy, t+dt).
The
Theoverall
overallarchitecture
architectureisisshown
shownbelow
below

Notice how it has three main blocks: Feature Encoder Block, Visual
Similarity Block, and an Iterative Update Block
Feature Extraction

RAFT performs feature extraction


on both input images using a
Convolutional Neural Network
(CNN) that consists of six
residual blocks and downsamples
each image to 1/8 resolution with
D feature maps.
Visual Similarity

The Visual Similarity is an all-pairs In the correlation volume, each pixel


Correlation Volume that relates the from feature map g¹ has a computed
pixels of I₁ to every single pixel of I₂ by correlation with every pixel from
computing the correlation of every feature map g², we call each of these
feature map at each pixel location correlations a 2D response map
Iterative Updates

The update block takes as inputs: the correlation features, current flow
estimate, context features, and the hidden features. Its architecture with
highlighted sub-blocks is displayed below.

The sub-blocks inside the Update Block are :

Feature Extraction Block — extracts motion features


from the correlation, flow, and I₁ (context network).
Recurrent Update Block — Recurrently computes flow
updates
Flow Head — Final convolutional layers that re-size
flow estimate to H/8 x W/8 x 2
Thank You
Action Item Point Person

Before we go 1
Add ideas from
2
Turn them into
3
Assign them to a
the flowchart here action items team member

Now that we've mapped out our


processes, we just have to Add top idea

figure out our next steps. And


then we're done!
Add top idea

Thanks for participating, and


Add top idea
see you in our next meeting!

BACK TO AGENDA PAGE


Use these free recolorable icons and
Resource Page illustrations in your Canva design.

Add text Add text Add text

Add text

Add text Add text


Add text
Flowchart

When workflows get complicated, a


flowchart can help. This visual map of
our processes can guide our actions,
identify roadblocks, and streamline
our activities for better efficiency.

LET'S BEGIN!

BACK TO AGENDA PAGE

You might also like