πŸ‘οΈ8,962
GitHubLinkedIn
Computer Vision and AI Applicationsβ€’β€’4 min readβ€’704 words

πŸ€– AI Conference - Poster Presentation

πŸ‘οΈ0reads (human + AI)πŸ€–0AI ingestions

πŸ€– AI Conference - Poster Presentation

This article summarizes a poster presentation on scaling trends for data poisoning in large language models (LLMs) presented at the AAAI conference. The presentation details research on the challenges of scaling data poisoning attacks against LLMs.

Key Points:

β€’ Explores the increasing vulnerability of LLMs to data poisoning as model size grows.

β€’ Investigates the scaling trends of different data poisoning techniques.

β€’ Presents mitigation strategies for reducing the impact of data poisoning.

πŸ”— Resources:

β€’ @RealAAAI β†— - AAAI Conference

Image

Image


πŸš€ Computer Vision - Shape of Motion Results

This article announces the release of the MegaSAM code and improved Shape of Motion results achieved using it. The research demonstrates significant progress in a challenging computer vision problem.

Key Points:

β€’ MegaSAM code is publicly available.

β€’ Updated Shape of Motion results show significant improvements.

β€’ Achieved progress previously deemed unattainable.

πŸ”— Resources:

β€’ Shape of Motion Results β†— - Project results

Image

Image


πŸ’‘ Computer Vision Research - CVPR Rejection and Community Concerns

This article discusses the rejection of a CVPR 2025 submission and raises concerns about potential repercussions for researchers voicing critical opinions within the research community.

Key Points:

β€’ A CVPR 2025 paper submission was rejected.

β€’ The author questions whether voicing concerns leads to negative consequences for researchers and their collaborators.

β€’ Highlights the importance of a fair and open research environment.


πŸ’‘ Academic Culture - Open Dialogue in Research

This article discusses the importance of open dialogue and the potential for negative consequences for researchers who express dissenting opinions within the research community.

Key Points:

β€’ A fair research community should not penalize researchers for voicing concerns.

β€’ Unhealthy academic environments may silence dissent.

β€’ The importance of fostering open and constructive criticism is stressed.


πŸš€ Book Launch - Prompt Engineering Book

This article announces the successful launch of a new prompt engineering book, highlighting strong initial sales figures.

Key Points:

β€’ A new prompt engineering book was launched.

β€’ 118 copies were sold within the first 24 hours.

β€’ The book's initial reception is positive.


✨ Academic Achievement - PhD Graduation

This article celebrates the graduation of Phillip Lippe with a PhD cum laude, highlighting his outstanding thesis and defense.

Key Points:

β€’ Phillip Lippe graduated with a PhD cum laude.

β€’ This achievement recognizes an outstanding thesis and defense.

β€’ The success is attributed to a strong advisor team.


πŸ€– Computer Vision - CVPR 2025 Acceptance

This article announces the acceptance of the TANGO paper (Training-free Embodied AI Agents for Open-world Tasks) to CVPR 2025.

Key Points:

β€’ TANGO paper accepted to CVPR 2025.

β€’ The success is attributed to the hard work of the student researchers.

β€’ The paper underwent multiple revisions.

πŸ”— Resources:

Image

Image


πŸš€ Border Security - Drone Surveillance Partnership

This article announces a partnership between Ondas and Volatus Drones to provide autonomous drone surveillance for border security.

Key Points:

β€’ Ondas and Volatus Drones are partnering on a border security project.

β€’ The solution integrates the Optimus System with scalable remote operations.

β€’ The goal is to provide 24/7 security and rapid response capabilities.

πŸ”— Resources:

β€’ Ondas Partnership Announcement β†— - Partnership details


πŸ€– Computer Vision - 3D Asset Generation

This article announces the acceptance of 3DTopia-XL to CVPR 2025, highlighting the open-sourced training code including the PrimX tokenization scheme.

Key Points:

β€’ 3DTopia-XL accepted to CVPR 2025.

β€’ Training code and PrimX tokenization scheme are open-sourced.

β€’ PrimX converts textured meshes into NxD tensors.

πŸ”— Resources:

β€’ 3DTopia-XL Code β†— - Project code

β€’ 3DTopia-XL Project Page β†— - Project details

Image

Image


πŸ€– Computer Vision - Image-to-Video Generation

This article announces the acceptance of Through-The-Mask (TTM) to CVPR 2025, an I2V generation framework leveraging mask-based motion trajectories.

Key Points:

β€’ Through-The-Mask (TTM) accepted to CVPR 2025.

β€’ TTM enhances object-specific motion and consistency in multi-object scenarios.

β€’ The framework uses mask-based motion trajectories.

πŸ”— Resources:

Image

Image


⭐️ Support

If you liked reading this report, please star ⭐️ this repository and follow me on Github β†—, 𝕏 (previously known as Twitter) β†— to help others discover these resources and regular updates.


Related Computer Vision and AI Applications Breakdowns

Drix10
Written by Drix10

Co founder @ PartPilot | 1 x Acquired Founder | Canopy @ f.inc | Cybersec @ DSU | 2x International Hackathon πŸ†. Read more on drix10.com.