Yux1angJi/DIFF
[ICASSP 2025] Diffusion Features to Bridge Domain Gap for Semantic Segmentation
This project helps computer vision engineers improve the accuracy of semantic segmentation models when applying them to new or different visual environments. It takes existing image datasets (like synthetic driving scenes) and trained segmentation models, then extracts and integrates 'universal features' from powerful text-to-image diffusion models. The result is a more robust segmentation model that performs better on real-world or novel datasets (like city driving or adverse weather conditions) without extensive re-training.
No commits in the last 6 months.
Use this if you need to deploy a semantic segmentation model trained on one type of image data to a visually distinct environment and want to maintain high accuracy without expensive re-annotation or retraining.
Not ideal if your segmentation task already performs well on its target domain, or if you are working with non-visual data.
Stars
18
Forks
—
Language
Python
License
—
Category
Last pushed
Nov 21, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/Yux1angJi/DIFF"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
cswry/SeeSR
[CVPR2024] SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution
JJLibra/SALAD-Pan
🤗 Official implementation for "SALAD-Pan: Sensor-Agnostic Latent Adaptive Diffusion for...
open-mmlab/mmgeneration
MMGeneration is a powerful toolkit for generative models, based on PyTorch and MMCV.
Janspiry/Image-Super-Resolution-via-Iterative-Refinement
Unofficial implementation of Image Super-Resolution via Iterative Refinement by Pytorch
hanjq17/Spectrum
[CVPR 2026] Adaptive Spectral Feature Forecasting for Diffusion Sampling Acceleration