refactor: move overview to 'projects'

nbs/index.qmd

@@ -3,6 +3,25 @@ title: home
 page-layout: article
 sidebar: false
 toc: false
+listing:
+  - id: projects-list
+    contents: "projects/*.qmd"
+    type: table
+    sort: "date desc"
+    date-format: iso
+    sort-ui: false
+    filter-ui: false
+    categories: false
+    fields: [title, date]
+  - id: posts-list
+    contents: "blog/posts"
+    type: table
+    sort: "date desc"
+    date-format: iso
+    sort-ui: false
+    filter-ui: false
+    categories: false
+    fields: [title, date]
 ---
 
 ## Hi, I'm Andrew Look.
@@ -13,39 +32,10 @@ You can learn more about me on [my "about" page](about/index.qmd), or you can ch
 
 ## Projects
 
-### Single-Line Machine Learning
-
-Lately I've been working on training ML models to generate single-line drawings in my style. I've open-sourced the code for my [models](https://github.com/andrewlook/singleline_models) and the code I used to prepare the [dataset](https://github.com/andrewlook/singleline_dataset).
-
-In part [part 1](./blog/posts/slml_part1.qmd), I cover how I'm using embeddings to filter my dataset of drawings. I also made an exploratory browser to visualize the embedding space of the drawings, and published it at [projector.andrewlook.com](https://projector.andrewlook.com). Here's a demo video:
-
-:::{.column-body-outset}
-{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/emb-projector.mp4" >}}
+:::{#projects-list}
 :::
 
-In [part 2](./blog/posts/slml_part2.qmd), I cover how I'm vectorizing the scans to prepare them for RNN/transformer training. I also wrote some fun preprocessing code to [join SVG paths](https://singleline-dataset.andrewlook.com/strokes.html) up into continuous lines. Here's a debug video I made to flip through part of my dataset and watch drawings go from multiple strokes (one color per stroke) to a single stroke:
-
-:::{.column-body-outset}
-{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/new_file_fps20.mp4" >}}
-:::
-
-In [part 3](./blog/posts/slml_part3.qmd), I cover a series of experiment results from training SketchRNN with different regularization and data augmentation, to see how it affects the results that the models can generate. Here are some of the resulting drawings:
-
-:::{.column-body-outset layout-ncol=3}
-{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/video-11.mp4" >}}
-
-{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/video-8.mp4" >}}
-
-{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/video-9.mp4" >}}
-:::
-
-In part 4 _(coming soon)_, I cover my experiments using a transformer model instead of an RNN, to see if the model can better handle long-term dependencies within the drawings.
-
-:::{.column-body-outset cap-location=bottom}
-![Architecture of the SketchFormer model.](https://i.ibb.co/rx1nbmt/sketchformer-architecture.png)
-:::
-
-
-
-
+## Posts
 
+:::{#posts-list}
+:::

nbs/projects/slml.qmd

@@ -0,0 +1,54 @@
+---
+title: SLML - Single-Line Machine Learning
+description: Overview of my project training ML models on a dataset of single-line drawings.
+categories: [singleline, machinelearning, slml]
+author: Andrew Look
+date: 2023-10-25
+date-modified: 2024-03-06
+image: https://i.ibb.co/Vj5kRw3/download.png
+comments:
+  utterances:
+    repo: andrewlook/notebooks
+cap-location: margin
+sidebar: true
+---
+
+# SLML - Single-Line Machine Learning
+
+Lately I've been working on training ML models to generate single-line drawings in my style. I've open-sourced the code for my [models](https://github.com/andrewlook/singleline_models) and the code I used to prepare the [dataset](https://github.com/andrewlook/singleline_dataset).
+
+## [Part 1 - Embedding Filtering](./blog/posts/slml_part1.qmd)
+
+In [part 1](./blog/posts/slml_part1.qmd), I cover how I'm using embeddings to filter my dataset of drawings. I also made an exploratory browser to visualize the embedding space of the drawings, and published it at [projector.andrewlook.com](https://projector.andrewlook.com). Here's a demo video:
+
+:::{.column-body-outset}
+{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/emb-projector.mp4" >}}
+:::
+
+## [Part 2 - Dataset Vectorization](./blog/posts/slml_part2.qmd)
+
+In [part 2](./blog/posts/slml_part2.qmd), I cover how I'm vectorizing the scans to prepare them for RNN/transformer training. I also wrote some fun preprocessing code to [join SVG paths](https://singleline-dataset.andrewlook.com/strokes.html) up into continuous lines. Here's a debug video I made to flip through part of my dataset and watch drawings go from multiple strokes (one color per stroke) to a single stroke:
+
+:::{.column-body-outset}
+{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/new_file_fps20.mp4" >}}
+:::
+
+## [Part 3 - Training SketchRNN](./blog/posts/slml_part3.qmd)
+
+In [part 3](./blog/posts/slml_part3.qmd), I cover a series of experiment results from training SketchRNN with different regularization and data augmentation, to see how it affects the results that the models can generate. Here are some of the resulting drawings:
+
+:::{.column-body-outset layout-ncol=3}
+{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/video-11.mp4" >}}
+
+{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/video-8.mp4" >}}
+
+{{< video "https://storage.googleapis.com/andrewlook-art-assets/andrewlook.com/videos/video-9.mp4" >}}
+:::
+
+## Part 4 - Training Sketchformer
+
+In part 4 _(coming soon)_, I cover my experiments using a transformer model instead of an RNN, to see if the model can better handle long-term dependencies within the drawings.
+
+:::{.column-body-outset cap-location=bottom}
+![Architecture of the SketchFormer model.](https://i.ibb.co/rx1nbmt/sketchformer-architecture.png)
+:::