@ErichLiang

PhD student @PrincetonVL. Previously at @Berkeley_AI, '21 Math/CS @Caltech. Interested in all things robotics and 3D computer vision. When in doubt, --force

erichliang.github.io/
Joined May 2021
  • Tweets28
  • Following130
  • Followers96
5Media
Dynamic camera intrinsics often appear in robotics and 3D vision—yet training and evaluation data for predicting them from RGB is surprisingly rare. Introducing InFlux++: our new real-world benchmark and synthetic training dataset for dynamic intrinsics prediction. 🧵1/6
1
18
2
131
10,846
Erich Liang retweeted
Arnold Caleb Asiimwe@acalebOct 12
1/ I’m excited to share our work on “Depth as Time in One-Step Generative Models”! alphaxiv.org/abs/2610.03626 TL;DR: We ask, when a model learns to generate an image in one forward pass, what happens to the denoising process? We observe that process unfolding over depth!
5
38
4
271
12,982
Erich Liang retweeted
Adithya Bhaskar@AdithyaNLPOct 5
The most profound paper of my PhD so far. We truly did something special to make LMs play/explain chess, needing both architectural & algorithmic innovation (natural-language analogue of the Alphazero algorithm) + applicable to many other domains. Please read on and share! 1/10
61
174
28
1,550
110,133
Erich Liang retweeted
Jingkang Yang@JingkangYSep 10
Hi #ECCV2026 — someone’s backpack got a MASSIVE upgrade 🎒 My ECCV status got an upgrade too: from PhD student to co-founder of @Action_Int_AI. Backpack? Now it’s my demo screen. First pitch: @CSProfKGD ✅ Next stop: everyone at Booth #44 — @Action_Int_AI × @rayneo_global
Kosta Derpanis
@CSProfKGD29 Sep 2024
Look who’s back! #ECCV2024
3
11
2
70
17,564
Introducing #Continuo. Intelligence that stays with you. At #ECCV: understand the poster in front of you, remember what matters, and find what to explore next. Chapter 01 / FOR HUMAN From Action Intelligence, building robotic foundation models. @eccvconf
9
46
59
120
6,215
I'll be at #ECCV2026; please DM / reach out if you'd like to chat! I'll also be presenting InFlux++ at: Main Poster Session 3: 📍ExHall #268, Fri Sep 11, 10:30–12:30 Privacy Preserving Computer Vision workshop: 📍ExHall #241 - 260, Wed Sep 9, 11:45–13:00 See you soon! 👋
Erich Liang@ErichLiangSep 1
Dynamic camera intrinsics often appear in robotics and 3D vision—yet training and evaluation data for predicting them from RGB is surprisingly rare. Introducing InFlux++: our new real-world benchmark and synthetic training dataset for dynamic intrinsics prediction. 🧵1/6
4
19
2,532
Erich Liang retweeted
Shmuel Berman@ShmuelBermanSep 2
Researchers are actively improving memory for LLMs/VLMs. But are we measuring it right? A system can score 100% by rereading the input for each question. If a human did that, we’d call it terrible memory! We test beyond accuracy with ECC: Efficiency, Compression, Calibration. 🧵
4
7
22
1,005
Evaluating SOTA methods on InFlux++ Real & InFlux reveals they still struggle with dynamic intrinsics. The good news? Finetuning AnyCalib on InFlux++ Synth consistently improves its focal length estimation across both benchmarks, showing the power of synthetic supervision. 🧵5/6
1
4
319
Want to benchmark your own dynamics intrinsics estimation method or train with InFlux++ Synth? The dataset, benchmark, code, videos, and live leaderboard are now available👇 🌐 Website: influx.cs.princeton.edu 📄 Paper: arxiv.org/abs/2607.05389 💻 Github: github.com/princeton-vl/InFl…
7
332
We have released an initial Infinigen2-Flying-Indoors dataset! It provides 2000 stereo videos, 24 frames each, with ground truth for depth, flow, normals, segmentation, albedo and more. Each group of 4 trajectories provides random synchronized views of the same procedural dynamic 3D scene. Code to replicate it will be released soon as infinigen2.0.0a2. See full youtube preview & huggingface link below ↓
2
8
2
40
15,763
1/ We've released Infinigen 2.0! Currently in preview. It creates indoor 3D scene files in 1min CPU time, and includes new and better materials --- all still fully procedural. Our new 2.0 design is highly efficient and allows easy control and recombination via Python APIs.
13
43
4
358
32,918
Erich Liang retweeted
Zhenjun Zhao@zhenjun_zhaoJul 7
InFlux++: Real and Synthetic Data for Estimating Dynamic Camera Intrinsics @ErichLiang, Caleb Kha-Uong, Chinmaya Saran, Sreemanti Dey, David W. Liu, Junhan Ouyang, Benjamin Zhou, @jiadeng tl;dr: synthetic video dataset + real-world benchmar arxiv.org/abs/2607.05389
6
19
1,640
Meet WAFT (Warping-Alone Field Transforms), our new optical-flow estimator. #1 on public benchmarks (Sintel & Spring), 1.3-4.1x faster than leading methods, and 2x lower memory. Key idea: replace cost volumes with high-res feature-space warping. Code and paper:👇
1
4
13
1,065
Erich Liang retweeted
Xindi Wu@cindy_x_wuJan 20
New #NVIDIA Paper We introduce Motive, a motion-centric, gradient-based data attribution method that traces which training videos help or hurt video generation. By isolating temporal dynamics from static appearance, Motive identifies which training videos shape motion in video generation. 🔗 research.nvidia.com/labs/sil… 1/10
12
123
10
641
145,577
Estimating camera intrinsics from video is key to 3D reconstruction, but most methods assume they’re fixed per video. What if the camera keeps zooming and refocusing? Meet InFlux, the first benchmark with per-frame ground truth for videos with dynamic intrinsics. 🧵1/5
1
4
1
15
1,618
Come check out our poster at #neurips2025!
Princeton Vision & Learning Lab
@PrincetonVL1 Dec 2025
Estimating camera intrinsics from video is key to 3D reconstruction, but most methods assume they’re fixed per video. What if the camera keeps zooming and refocusing? Meet InFlux, the first benchmark with per-frame ground truth for videos with dynamic intrinsics. 🧵1/5
4
333
Princeton365: A Diverse Dataset with Accurate Camera Pose Karhan Kayan, Stamatis Alexandropoulos, Rishabh Jain, Yiming Zuo, @ErichLiang, @jiadeng tl;dr: dataset with gt camera pose and 6 DoF camera motion; new optical-flow based metric arxiv.org/abs/2506.09035
5
28
1,934
Load more