New England Computer Vision (NECV) Workshop 2024

Yale University, New Haven, CT

Friday, November 22, 2024

---

The New England Computer Vision Workshop (NECV) brings together researchers in computer vision and related areas for an informal exchange of ideas through a full day of presentations and posters. Held conveniently after the CVPR deadline and before the NeurIPS conference, NECV offers opportunities to network and showcase research. NECV attracts researchers from universities and industry research labs in New England. As in previous years, the workshop will focus on graduate student presentations. Welcome to Yale!

- Alex Wong

---

Registration and Submission

Academic researchers: Participation is free for all researchers at academic institutions. Please register here and submit your abstract here.

Industry participants: For our industry friends, a limited number of registrations are available for a fee. Please register here.

Deadlines: Early-bird registration (lunch provided) by November 7. Please register by November 15 and submit by November 17. Oral decisions will be released by November 19.

Submission guidelines: Please submit a one-page PDF abstract using the CVPR 2025 rebuttal template. Please include the title of your work and the list of authors in the abstract. You may present work that has already been published or work that is in progress. All relevant submissions will be granted a poster presentation, and selected submissions from each institution will be granted 8-minute oral presentations. Post-docs and faculty may submit for poster presentations, but oral presentations are reserved for graduate students. There will be no publications resulting from the workshop, so presentations will not be considered "prior peer-reviewed work" according to any definition we are aware of. Thus, work presented at NECV can be subsequently submitted to other venues without citation. The workshop is after the CVPR submission deadline, so come and show off your new work in a friendly environment. It's also just before the NeurIPS conference, so feel free to come and practice your presentation.

---

Presentation

Oral presentation: Each presentation is allocated a 6-minute slot, with an additional 2 minutes dedicated to questions. We kindly request all oral presenters to bring their machines for their presentation. The presentation equipment supports both HDMI and Type-C for screen sharing. Please arrive at least 5 minutes before the scheduled oral session to test your machine and ensure compatibility with the provided equipment. Similar to regular conferences, we have also allocated poster boards for oral presenters. Please find your poster ID.

Poster presentation: Please locate the correct poster board to display your poster. Easels and foam cores will be provided for mounting posters, accommodating sizes up to 36x48 inches. The foam cores are not attached, allowing flexibility for landscape or portrait orientation. You are welcome to use any format within that size limit.

---

Logistics

Schedule

Time	Topic
9:00-10:00	Registration & Poster Setup
10:00-10:10	Welcome & Opening
10:15-11:15	Oral Session I
[10:15] Learning to Edit Visual Programs with Self-Supervision [10:25] What if Eye? Computationally Emulating the Evolution of Visual Intelligence [10:35] Orient Anything [10:45] Score Distillation via Reparameterized DDIM [10:55] Straightening Flow Matching Models by Learning Interpolants [11:05] Time of the Flight of the Gaussians:Fast and Accurate Dynamic Time-of-Flight Radiance Fields
11:25-12:25	Poster Session I
[1] Augundo: Scaling up augmentations for monocular depth completion and estimation [2] Orient Anything [3] What if Eye? Computationally Emulating the Evolution of Visual Intelligence [4] Enhancing Autonomous Navigation by Imaging Hidden Objects using Single-Photon LiDAR [5] Active Appearance and Spatial Variation Can Improve Visibility in Area Labels for Augmented Reality [6] E-BARF: Bundle Adjusting Neural Radiance Fields from a Moving Event Camera [7] CUTS: A Deep Learning and Topological Framework for Multigranular Unsupervised Medical Image Segmentation [8] GigaHands: A Massive Annotated Dataset of Bimanual Hand Activities [9] Real-Time Temporally Consistent Depth Completion for VR-Teleoperated Robots [10] OneGaze: A Unified Model for Estimating Gazeing Egocentric Videos and Still Images [11] Audio Geolocation: An Investigation with Natural Sounds [12] Straightening Flow Matching Models by Learning Interpolants [13] CP-TRPCA: A Novel Approach to Robust Tensor PCA [14] ODTFormer: Efficient Stereo-based Obstacle Detection with Deformable Matching Cost Attention [15] Text-Aware Diffusion for Policy Learning [16] Hybrid CNN with Multimodal Data for Early Alzheimer's Disease Forecasting [17] Combining Observational Data and Language for Species Range Estimation [18] Learning to Edit Visual Programs with Self-Supervision [19] Time of the Flight of the Gaussians:Fast and Accurate Dynamic Time-of-Flight Radiance Fields [20] Personalized Representation from Personalized Generation [21] Audio-Visual Speech Separation via Bottleneck Iterative Network [22] Score Distillation via Reparameterized DDIM [23] Pre-trained Vision-Language Models Learn Discoverable Visual Concepts [24] Adaptive Correspondence Scoring for Unsupervised Medical Image Registration [25] Mine yOur owN Anatomy: Revisiting Medical Image Segmentation with Extremely Limited Labels [26] ProtoDepth: Unsupervised Continual Depth Completion with Prototypes [27] RSA: Resolving scale ambiguities in monocular depth estimators through language descriptions
12:30-1:30	Lunch
1:50-2:40	Oral Session II
[1:50] Event fields: Capturing light fields at high speed, resolution, and dynamic range [2:00] RICH: Non-stochastic Robust Inference of Camera Headings [2:10] DELTA: Dense Efficient Long-range 3D Tracking for any video [2:20] Smooth Spline: Provably Smoothing the Decision Manifolds of Deep Neural Networks After Training [2:30] The GAN is dead; long live the GAN! A Modern GAN Baseline [2:40] The iNaturalist Sounds Dataset
2:55-3:55	Poster Session II
[1] WildSAT: Learning Satellite Image Representations from Wildlife Observations [2] Monocular Dynamic Gaussian Splatting is Fast and Brittle but Smooth Motion Helps [3] ImageFlowNet: Forecasting Multiscale Image-Level Trajectories of Disease Progression with Irregularly-Sampled Longitudinal Medical Images [4] RICH: Non-stochastic Robust Inference of Camera Headings [5] HouseCrafter: Lifting Floorplans to 3D Scenes with 2D Diffusion Models [6] Test-time adaptation for depth completion [7] Smooth Spline: Provably Smoothing the Decision Manifolds of Deep Neural Networks After Training [8] Differentiable Robot Rendering [9] All-day Depth Completion [10] ACDIT: Interpolating Autoregressive Conditional Modeling and Diffusion Transformer [11] DiffKillR: Killing and Recreating Diffeomorphisms for Cell Annotation in Microscopy Images [12] Benchmarking Single-Positive Multi-Label Methods using the L48S Dataset [13] UnCLe: Unsupervised Continual Learning of Depth Completion [14] The iNaturalist Sounds Dataset [15] Solving New Tasks by Adapting Internet Video Knowledge [16] Evidential Neural Radiance Fields [17] Event fields: Capturing light fields at high speed, resolution, and dynamic range [18] As-Ortho-As-Possible View Selection for Decomposing Appearance in Scene Reconstruction [19] Similarity Group Equivariant Convolutional Networks [20] Zero-Shot Monocular Scene Flow Estimation in the Wild [21] Vessel-aware aneurysm detection using multi-scale deformable 3D attention [22] DELTA: Dense Efficient Long-range 3D Tracking for any video [23] Studying Priming Effect in Vision-Language Models [24] Thinking in Space: How Multimodal Large Language Models See, Remember and Recall Spaces [25] Squeezing Water from a Stone: Improving Pre-Trained Self-Supervised Embeddings Through Effective Entropy Maximization [26] Compositional Content Recommendation and Controllable Object Placement with Multimodal LLM [27] iNatator: Obtaining Expert Feedback on Species Ranges [28] NeST: Neural Stress Tensor Tomography by leveraging 3D Photoelasticity [29] How Can Objects Help Video-Language Understanding? [30] Token-Shuffle: Towards High-Resolution Image Generation with Autoregressive Models [31] Image Reconstruction from Readout-Multiplexed Single-Photon Detector Arrays [32] The GAN is dead; long live the GAN! A Modern GAN Baseline [33] Generate, Transduct, Adapt: Iterative Transduction with VLMs
3:55-4:25	Coffee Break
4:30-4:40	Statement by Jeffrey Brock, Dean of Yale SEAS
4:40-6:10	Oral Session III
[4:40] NeST: Neural Stress Tensor Tomography by leveraging 3D Photoelasticity [4:50] Image Reconstruction from Readout-Multiplexed Single-Photon Detector Arrays [5:00] Differentiable Robot Rendering [5:10] ACDIT: Interpolating Autoregressive Conditional Modeling and Diffusion Transformer [5:20] Vessel-aware aneurysm detection using multi-scale deformable 3D attention [5:30] Squeezing Water from a Stone: Improving Pre-Trained Self-Supervised Embeddings Through Effective Entropy Maximization [5:40] Solving New Tasks by Adapting Internet Video Knowledge [5:50] Generate, Transduct, Adapt: Iterative Transduction with VLMs [6:00] ImageFlowNet: Forecasting Multiscale Image-Level Trajectories of Disease Progression with Irregularly-Sampled Longitudinal Medical Images
6:10-6:20	Closing Remarks

Venue

The workshop will be held at Marsh Auditorium, Yale Science Building, Yale University.

WiFi

Log on to "YaleGuest" and accept the terms.

Hotel Accommodation

• Hotel Marcel
• Hampton Inn & Suites New Haven - South - West Haven
• Homewood Suites by Hilton - Orange / New Haven
• Courtyard by Marriott New Haven Orange/Milford
• Residence Inn by Marriott New Haven / Hamden

Contact

For any questions related to our upcoming workshop, please contact our committee at necv2024yale@gmail.com.

---

Organizers

Host: Alex Wong.

Program committee: Hyoungseob Park, Ziyao Zeng, Daniel Wang, Patrick Rim, Younjoon Chung, Ruxiao Duan, Xiaoran Zhang, Vadim Ezhov, Nick He, Xien Chen, Rit Gangopadhyay, and Jingcheng Ni.

Logistics committee: Hyoungseob Park, Vadim Ezhov, and Rit Gangopadhyay.

Corporate relations committee: Patrick Rim, Runjian Chen, Nick He, and Xien Chen.

Website chair: Ruxiao Duan.

Steering committee: Subhransu Maji (UMass Amherst), Erik Learned-Miller (UMass Amherst), Kate Saenko (Boston University), Yun (Raymond) Fu (Northeastern University), Octavia Camps (Northeastern University), Todd Zickler (Harvard), James Tompkin (Brown), Benjamin Kimia (Brown), Phillip Isola (MIT), Pulkit Agrawal (MIT), SouYoung Jin (Dartmouth), Adithya Pediredla (Dartmouth), and Yu-Wing Tai (Dartmouth).

---

Acknowledgements

We thank Samson Timoner for helping us arrange NECV 2024.