Research map / Video / Video segmentation and tracking
Recycle-GAN: Unsupervised Video Retargeting
Aayush Bansal, Shugao Ma, Deva Ramanan, Yaser Sheikh
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add recycle (temporal) loss in place of cycle loss alone (ablation) | Cycle-GAN · Viper image-to-labels (all) | mean pixel accuracy (%): 35.5 -> 56.0 | Improved |
| add recycle (temporal) loss in place of cycle loss alone (ablation) | Cycle-GAN · Viper image-to-labels (all) | average class accuracy (%): 7.1 -> 11.3 | Improved |
| add recycle (temporal) loss in place of cycle loss alone (ablation) | Cycle-GAN · Viper image-to-labels (all) | IoU (%): 4.2 -> 8.2 | Improved |
| combine cycle loss and recycle loss versus recycle loss alone (ablation) | Recycle-GAN · Viper image-to-labels (all) | IoU (%): 8.2 -> 8.9 | Improved |
| combine cycle loss and recycle loss versus recycle loss alone (ablation) | Recycle-GAN · Viper image-to-labels (all) | mean pixel accuracy (%): 56.0 -> 53.7 | Hurt |
| combine cycle loss and recycle loss versus recycle loss alone (night) (ablation) | Recycle-GAN · Viper image-to-labels (night) | mean pixel accuracy (%): 45.2 -> 33.7 | Hurt |
| combine cycle loss and recycle loss versus recycle loss alone (ablation) | Recycle-GAN · Viper labels-to-image (all) | normalized FCN score: 0.39 -> 0.48 | Improved |
| use recycle loss instead of cycle loss (ablation) | Cycle-GAN · Viper labels-to-image (all) | normalized FCN score: 0.30 -> 0.39 | Improved |
| use recycle loss instead of cycle loss (day) (ablation) | Cycle-GAN · Viper labels-to-image (day) | normalized FCN score: 0.33 -> 0.33 | No clear effect |
| use recycle loss instead of cycle loss (rain) (ablation) | Cycle-GAN · Viper labels-to-image (rain) | normalized FCN score: 0.39 -> 0.37 | No clear effect |
| use spatiotemporal recycle loss instead of cycle loss, human realism judgment (ablation) | Cycle-GAN · Human study real vs generated | generated videos judged real (%): 7.3% -> 28.3% | Improved |
| use spatiotemporal recycle loss instead of cycle loss, pairwise preference (ablation) | Cycle-GAN · Human study pairwise naturalness | chosen as more realistic (%): 8% -> 76% | Improved |
Later papers that compared with this method
1 later accepted paper reports Recycle-GAN as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Learning Temporally and Semantically Consistent Unpaired Video-to-Video Translation through Pseudo-Supervision from Synthetic Optical Flow (AAAI 2022 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RIGID: Recurrent GAN Inversion and Editing of Real Face Videos (ICCV 2023 · similarity 0.85)
- Video-to-Video Synthesis (NeurIPS 2018 · similarity 0.84)
- RIGID: Recurrent GAN Inversion and Editing of Real Face Videos and Beyond (IJCV 2025 · similarity 0.84)
- Video Repurposing from User Generated Content: A Large-scale Dataset and Benchmark (AAAI 2025 · similarity 0.84)
- Diverse Generation from a Single Video Made Possible (ECCV 2022 · similarity 0.84)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · similarity 0.82)
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.
