Work    Notes    Info

AI Lounge Chair 02
(Silhouette from AI)


2024
For Salone Satellite 2024


Stainless steel frame with rope seat and back.
Photography by Lee Jin Woo & Studio GYU



This is the second AI Lounge Chair. I focused more on curating the AI dataset, creating around 250 sketches. After AI training and generation, I selected some ambiguous, unique-shaped images from the results. This resulted in a lounge chair that deviates from typical usage.

For participation in the <SALONESATELLITE 2024> exhibition, all structures were designed to be disassembled into minimal units for transportation with reduced volume.









Algorithm: Denoising Diffusion Probabilistic Models (DDPM)
Architecture: UNet2DModel
Training Type: Unconditional

Image Preprocessing:
- 209 Hand Sketches
- Flip All Images
- Add Rescaled Images
- Resize to 256 x 256
- Random Crop

Hyperparameters:
- Batch Size: 16
- Epochs: 220
- Learning Rate: 1e-4 (0.0001)

Optimization:
- Gradient Accumulation Steps: 1
- Mixed Precision: fp16

Image Generation:
- Device: CUDA (GPU)
- Number of inference steps: 1000
- Batch size for generation: 16

Final Image: Performed inference at 160 epochs
© 2025 HyungGyu Rhee