Research map / Image generation / Editing and style transfer
TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use transformation-vector preservation in a learned siamese space instead of cycle-consistency | TraVeLGAN · Abacus and crossword; volcano and jack-o-lantern; clock and hourglass; toucan and rock beauty | FID: 1.221 -> 1.026; 0.527 -> 0.032; 0.727 -> 0.698; 0.381 -> 0.206 | Helped |
| Use transformation-vector preservation in a learned siamese space instead of cycle-consistency | TraVeLGAN · Abacus and crossword; volcano and jack-o-lantern; clock and hourglass; toucan and rock beauty | Discriminator score: 0.014 -> 0.035; 0.191 -> 0.206; 0.066 -> 0.074; 0.064 -> 0.145 | Helped |
| Use transformation-vector preservation instead of cycle-consistency | TraVeLGAN · Apples to oranges; Van Gogh to landscape; Ukiyoe to landscape; CelebA glasses; CelebA hats | SSIM: 0.424 -> 0.302; 0.327 -> 0.183; 0.260 -> 0.222; 0.608 -> 0.499; 0.437 -> 0.420 | Mixed |
| Use transformation-vector preservation instead of cycle-consistency | TraVeLGAN · Sketches and shoes | Pixel MSE: 0.047 -> 0.060; 0.148 -> 0.267 | Hurt |
| Reduce latent-space size to 100 (ablation) | TraVeLGAN · Sketches and shoes | Pixel MSE: 0.060 -> 0.069; 0.267 -> 0.370 | Hurt |
| Increase latent-space size to 2000 (ablation) | TraVeLGAN · Sketches and shoes | Pixel MSE: 0.060 -> 0.064; 0.267 -> 0.274 | Mixed |
| Preserve pairwise distances in pixel space rather than using learned latent-space transformations | DistanceGAN · Toucan and rock beauty | FID | Hurt |
Closest papers by meaning
- Unsupervised Image-to-Image Translation Networks (NeurIPS 2017 · similarity 0.93)
- DeepI2I: Enabling Deep Hierarchical Image-to-Image Translation by Transferring from GANs (NeurIPS 2020 · similarity 0.92)
- Rethinking the Paradigm of Content Constraints in Unpaired Image-to-Image Translation (AAAI 2024 · similarity 0.90)
- Compressing Image-to-Image Translation GANs Using Local Density Structures on Their Learned Manifold (AAAI 2024 · similarity 0.90)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 2024 · similarity 0.89)
- Unlocking Pre-trained Image Backbones for Semantic Image Synthesis (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.