Skip to content

Commit

Permalink
Update index.html
Browse files Browse the repository at this point in the history
  • Loading branch information
@yashdave003
yashdave003 committed Nov 20, 2024
1 parent 5a99e76 commit c766615
Showing 1 changed file with 151 additions and 5 deletions.
156 changes: 151 additions & 5 deletions 5/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -127,6 +127,8 @@
<h1>CS 180 Project: Diffusion Models and Image Processing</h1>
<h3>Table of Contents</h3>

<!-- Part A -->
<h4>Part A: Implementation and Analysis</h4>
<ul class="actions">
<li><a href="#setup" class="arrow scrolly"><span class="label">Part 0: Setup</span></a></li>

Expand All @@ -148,6 +150,31 @@ <h3>Table of Contents</h3>
<li><a href="#hybrid" class="arrow scrolly">1.9 Hybrid Images</a></li>
</ul>
</ul>

<!-- Part B -->
<h4>Part B: Model Training and Development</h4>
<ul class="actions">
<li><a href="#single-step" class="arrow scrolly"><span class="label">Part 1: Training a Single-Step Denoising UNet</span></a></li>
<ul class="nested-list">
<li><a href="#unet-impl" class="arrow scrolly">1.1 Implementing the UNet</a></li>
<li><a href="#unet-training" class="arrow scrolly">1.2 Using the UNet to Train a Denoiser</a></li>
<ul class="nested-list">
<li><a href="#training" class="arrow scrolly">1.2.1 Training</a></li>
<li><a href="#ood-testing" class="arrow scrolly">1.2.2 Out-of-Distribution Testing</a></li>
</ul>
</ul>

<li><a href="#diffusion-training" class="arrow scrolly"><span class="label">Part 2: Training a Diffusion Model</span></a></li>
<ul class="nested-list">
<li><a href="#time-conditioning" class="arrow scrolly">2.1 Adding Time Conditioning to UNet</a></li>
<li><a href="#unet-training-2" class="arrow scrolly">2.2 Training the UNet</a></li>
<li><a href="#unet-sampling" class="arrow scrolly">2.3 Sampling from the UNet</a></li>
<li><a href="#class-conditioning" class="arrow scrolly">2.4 Adding Class-Conditioning to UNet</a></li>
<li><a href="#class-sampling" class="arrow scrolly">2.5 Sampling from the Class-Conditioned UNet</a></li>
</ul>

<li><a href="#bells-whistles" class="arrow scrolly"><span class="label">Bells and Whistles</span></a></li>
</ul>
</header>
</section>

Expand Down Expand Up @@ -418,29 +445,148 @@ <h4>1.7.3 Text-Conditional Image-to-image Translation</h4>

<section id="anagrams" class="section">
<h3>1.8 Visual Anagrams</h3>
<p>
The first two pictures were the best visual anagrams I got when combining the prompts, "an oil painting of an old man" and "an oil painting of people around a campfire".
The 3rd picture I have included for fun because the anagram is fully right side up - I think both aspects of the picture are quite clear and meld together surprisingly well.
</p>
<div class="image-container">

<div class="image-box">
<img src="media\man_campfire_1.jpg" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Original </div>
<div class="caption"> Old Man + Campfire Anagram </div>
</div>
<div class="image-box">
<img src="media\man_campfire_2.jpg" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Original </div>
<div class="caption"> Old Man + Campfire Anagram </div>
</div>

<div class="image-box">
<img src="media\man_campfire.jpg" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Original </div>
<div class="caption"> Old Man + Campfire Right-side Up </div>
</div>
</div>
<p>
I chose to incorporate dogs into sceneries after playing around with some other combinations. Making anagrams combining two sceneries sometimes proved to be difficult, as did combining weaker prompts (eg. a photo of a man, a photo of a hipster barista) with landscapes that tended to dominate. The ones below were the best as the dogs are strikingly clear, especially the second one.
For the third bonus anagram, it honestly felt like both the coast and rocket were quite prominent even without flipping it over so I chose to include it.
</p>
<div class="image-container">
<div class="image-box">
<img src="media\snowy_dog_anagram.png" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Snowy Village + Dog Anagram </div>
</div>
<div class="image-box">
<img src="media\amalfi_dog_anagram.png" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Amalfi Coast + Dog Anagram </div>
</div>
<div class="image-box">
<img src="media\amalfi_rocket.jpg" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Amalfi Coast + Rocket Anagram </div>
</div>
</div>

</section>

<section id="hybrid" class="section">
<h3>1.9 Hybrid Images</h3>
<!-- Content will be added here -->
<div class="image-container">
<div class="image-box">
<img src="media\skull_waterfall_hybrid_1.png" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Skull (Far) + Waterfall (Close) </div>
</div>
<div class="image-box">
<img src="media\skull_waterfall_hybrid_3.png" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Skull (Far) + Waterfall (Close) </div>
</div>
<div class="image-box">
<img src="media\skull_waterfall_hybrid_2.png" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Skull (Far) + Waterfall (Close) </div>
</div>
</div>

<div class="image-container">
<div class="image-box">
<img src="media\dog_campfire_hybrid.png" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Campfire (Close) + Dog (Far) </div>
</div>
<div class="image-box">
<img src="media\rocket_campfire_hybrid_1.png" alt="Output with num=10" class="enlarged-image">
<div class="caption"> Campfire (Close) + Rocket (Far) </div>
</div>
</div>
</section>

<!-- Part B Sections -->
<section id="single-step" class="section">
<h2>Part 1: Training a Single-Step Denoising UNet</h2>

<section id="unet-impl" class="section">
<h3>1.1 Implementing the UNet</h3>
<div class="image-box">
<img src="media\fig1.jpg">
</div>
</section>

<section id="unet-training" class="section">
<h3>1.2 Using the UNet to Train a Denoiser</h3>

<div class="image-box">
<img src="media\fig3_noise_on_mnist.png">
</div>

<section id="training" class="section">
<h4>1.2.1 Training</h4>
<div class="image-box">
<img src="media\fig4_training_loss_curve.png">
</div>
<div class="image-box">
<img src="media\fig5_test_epoch_1.png">
</div>
<div class="image-box">
<img src="media\fig6_test_epoch_5.png">
</div>
</section>

<section id="ood-testing" class="section">
<h4>1.2.2 Out-of-Distribution Testing</h4>
<div class="image-box">
<img src="media\fig7_ood.png">
</div>
</section>
</section>
</section>

<section id="diffusion-training" class="section">
<h2>Part 2: Training a Diffusion Model</h2>

<section id="time-conditioning" class="section">
<h3>2.1 Adding Time Conditioning to UNet</h3>
<!-- Content will be added here -->
</section>

<section id="unet-training-2" class="section">
<h3>2.2 Training the UNet</h3>
<!-- Content will be added here -->
</section>

<section id="unet-sampling" class="section">
<h3>2.3 Sampling from the UNet</h3>
<!-- Content will be added here -->
</section>

<section id="class-conditioning" class="section">
<h3>2.4 Adding Class-Conditioning to UNet</h3>
<!-- Content will be added here -->
</section>

<section id="class-sampling" class="section">
<h3>2.5 Sampling from the Class-Conditioned UNet</h3>
<!-- Content will be added here -->
</section>
</section>

<section id="bells-whistles" class="section">
<h2>Bells and Whistles</h2>
<!-- Content will be added here -->
</section>
</section>
</div>

Expand Down

0 comments on commit c766615

Please sign in to comment.