In our CVPR'25 paper, we introduced the ViCaS dataset which contains 20,000+ videos with both detailed video captions, as well as pixel-precise masks for selected objects with phrase-grounding (1/4)
1
2
428
For those of you looking to extend their Video-LLMs with spatial intelligence capability, this dataset is a potential game-changer. ViCaS is the largest, human-annotated video dataset that provides both captions as well as grounded segmentation masks (3/4)
1
1
63
🌍 Webpage: ali2500.github.io/vicas-proj…
📜 Arxiv: arxiv.org/abs/2412.09754
🧑💻Github: github.com/Ali2500/ViCaS
🤗HuggingFace: huggingface.co/datasets/Ali2…
(4/4)
1
52
Ali Athar retweeted
I'm happy to share our latest work "MaskBit: Embedding-free Image Generation via Bit Tokens" which tackles class-conditional image generation with sota results!
➡️Project page: weber-mark.github.io/project…
➡️preprint: arxiv.org/pdf/2409.16211
Ali Athar retweeted
📣📣 Hiring a PhD-Intern 📣📣
Work with me on Dynamic 3D Gaussians at the Meta Boston office for 6 months in summer 2025!
Apply here: metacareers.com/jobs/1054974…
+ write me your questions / link your most relevant work via email or twitter.
Dynamic 3D Gaussians: Tracking by Persistent Dynamic View Synthesis
dynamic3dgaussians.github.io…
We model the world as a set of 3D Gaussians that move & rotate over time. This extends Gaussian Splatting to dynamic scenes, with accurate novel-view synthesis and dense 3D trajectories.
Ali Athar retweeted
Check out our work on fine-tuning of image-conditional diffusion models for depth and normal estimation.
Widely used diffusion models can be improved with single-step inference and task-specific fine-tuning, allowing us to gain better accuracy while being 200x faster!⚡
🧵(1/6)
Ali Athar retweeted
It's been a pleasure to work on this with @GerardPonsMoll1
I think we found a really effective formulation for training large-scale strong pose and shape models.
Here are some more qualitative results on tough, in-the-wild YouTube dance videos.
For 3D pose some use different keypoints, others SMPL and other models. It's a mess!
With Neural Localizer Fields, we can choose the output at test time! allowing to train using any. Results are real time and SOTA across the board.
arxiv.org/pdf/2407.07532
@Istvan_Sarandi
Ali Athar retweeted
Happy to share that #PointVOS has been accepted to #CVPR2024 🎉 It was a great collaboration with Sabarinath Mahadevan, Paul Voigtlaender, Bastian Leibe🥳 I will be in Seattle next week to present our poster😊
📜Paper: arxiv.org/pdf/2402.05917
🌐 Website: pointvos.github.io
Successfully defended my PhD at the @RWTHVisionLab! I'm thankful to my supervisor, colleagues, family members and to God for this incredible 5-year experience! Aside from the professional/research experience, I'll cherish the personal bonds I made here for a long time to come.
As one journey ends, another beings! For the next phase in life, I've moved to the Bay Area and taken up a Research Scientist position at @BytedanceTalk where I'll continue to work on exciting research problems related to video understanding.
193
The poster for our #CoRL2023 paper on Lidar 4D panoptic segmentation will be up on Thursday (Nov 9) at 4:15pm! We propose a cool new multimodal approach for Lidar segmentation + tracking that leverages both Lidar scan and camera image data.
Website: waabi.ai/4d-former/
Ali Athar retweeted
🎉I’m pleased to announce our #ICCV2023 paper: " 💣DynaMITe: Dynamic Query Bootstrapping for Multi-object Interactive Segmentation Transformer."
Paper: arxiv.org/abs/2304.06668
Project: amitrana001.github.io/DynaMI…
Code: github.com/amitrana001/DynaM…
@Pandoro89 @RWTHVisionLab
Ali Athar retweeted
Dynamic 3D Gaussians: Tracking by Persistent Dynamic View Synthesis
dynamic3dgaussians.github.io…
We model the world as a set of 3D Gaussians that move & rotate over time. This extends Gaussian Splatting to dynamic scenes, with accurate novel-view synthesis and dense 3D trajectories.
I arrived to put up my @CVPR publication poster and saw that the poster from the previous session was still there. I was slightly annoyed but then I read this note pasted on the side and immediately felt guilty and sad.
💥 I'm pleased to announce our @CVPR Highlight paper: "TarViS: A Unified Approach for Target-based Video Segmentation".
💥TL;DR: We developed a single, jointly trained model that can tackle multiple video segmentation tasks e.g. VOS, VIS, VPS and achieve SOTA performance.
💥I'm excited to present our work at @CVPR in Vancouver in a couple of weeks! Our poster session is in the Thursday AM slot (10:30-12:00) on Jun 22.
💥YouTube: youtube.com/watch?v=qecFRmSY…
💥PDF: arxiv.org/pdf/2301.02657.pdf
💥Code + trained models: github.com/Ali2500/TarViS
1
4
308
I'm pleased to announce our new dataset for object segmentation in video called BURST (Benchmark for Unified Recognition, Segmentation and Tracking).
URL: github.com/Ali2500/BURST-ben…
This'll be a game-changer for people working on VIS, VOS, MOTS & open-world tracking: (1/5)












