FelTris/durf
Code release for my thesis 'Neural Rendering for Dynamic Urban Scenes'. We use Neural Radiance Fields to perform novel-view-synthesis in unbounded outdoor scenes and jointly regress 3D bounding box poses.
This project helps self-driving car developers and researchers create realistic simulations of urban environments. It takes real-world sensor data (like images and LiDAR from datasets like Waymo) or simulated data (from CARLA) and generates new, never-before-seen views of dynamic urban scenes. The output is a highly realistic 3D model that allows you to observe the scene from any camera angle, including the ability to manipulate individual objects.
No commits in the last 6 months.
Use this if you need to generate diverse and high-fidelity visual data for training and testing self-driving car perception and decision-making systems from existing real-world or simulated sensor recordings.
Not ideal if you primarily need to generate static, non-urban scenes or if your main goal is 2D image synthesis rather than 3D scene reconstruction and novel view generation.
Stars
40
Forks
2
Language
Python
License
—
Category
Last pushed
Nov 04, 2022
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/FelTris/durf"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
weiyithu/NerfingMVS
[ICCV 2021 Oral] NerfingMVS: Guided Optimization of Neural Radiance Fields for Indoor Multi-view Stereo
Jumpat/SegmentAnythingin3D
Segment Anything in 3D with NeRFs (NeurIPS 2023 & IJCV 2025)
POSTECH-CVLab/SCNeRF
[ICCV21] Self-Calibrating Neural Radiance Fields
half-potato/nmf
Our method takes as input a collection of images (100 in our experiments) with known cameras,...
synthesiaresearch/humanrf
Official code for "HumanRF: High-Fidelity Neural Radiance Fields for Humans in Motion"