Papers

Detecting Photoshopped faces by scripting Photoshop

Creates 3D character animation from a single image.

Easily find similar images.

Image inpainting for irregular holes using partial convolutions

Removes noise from images.

Turns ordinary video into slow-motion video.

Outputs a photorealistic video that accurately depicts the content of the input video.

Deep learning-based, fast multi-object tracker

A polyhedral framework for 3D multi-object tracking

AI that creates a 3D model from a single 2D photo

AI that creates a 3D model from a single photo of a person

Text region segmentation for OCR.

AI Toolkit for Healthcare Imaging

Canonical 3D pose networks for non-rigid structure from motion

Region selection -> object selection, good for use as an annotation tool.

Virtual walks in Google Street View using PoseNet and applying Deep Learning models to recognize actions.

Github - virtual-walk

Unsupervised Learning of visual features by contrasting cluster assignments

Probably needed for tracking multiple people

A practical AI model that detects cracked areas on concrete surfaces using alibi-detect's VAE

See Alibi Detect#Surface Crack Detection.

mist over green hills -> DeepDaze_-_Mist_over_green_hills.jpg

Generating images from text.

Simple command line tool for text to image generation using OpenAI's CLIP and Siren (Implicit neural representation network)

See Deep Daze.

Action recognition with skeletons

See MMSkeleton.

Predicts depth maps from images.

See Depth Prediction, Monodepth2.

Applies ML to skeletons from OpenPose. 9 actions; multiple people.
(Warning: unfortunately, this is only useful for a course demo, not for real-world applications

https://github.com/felixchenfy/Realtime-Action-Recognition

See Skeleton Based Action Recognition.

A joint deep network for feature line detection and description.

https://github.com/cvg/SOLD2

See Self-supervised Occlusion-aware Line Description and Detection (SOLD2).

PyTorch implementation and pretrained models for DINO.

https://github.com/facebookresearch/dino

See Self-Supervised Vision Transformers with DINO.

Easy face/hand/pose recognition on the web

https://github.com/midiblocks/handsfree

Visually probe the behavior of trained machine learning models with minimal coding.

https://pair-code.github.io/what-if-tool/

https://github.com/PAIR-code/what-if-tool

Physics-based human motion estimation and synthesis from videos - Physics-based Human Motion Estimation and Synthesis from Videos (ICCV 2021)

https://nv-tlabs.github.io/physics-pose-estimation-project-page/

Detector-free local feature matching with transformers

https://github.com/zju3dv/LoFTR

Erasing objects from high-resolution images

https://github.com/saic-mdal/lama

Self-supervised geometric correspondence for category-level 6D object pose estimation

https://kywind.github.io/self-pose

Extracts depth information from video for street views

• https://arxiv.org/abs/2303.00749
• https://ziyang-xie.github.io/s-nerf/

Open source for animating children's drawings

Alias: sam-hq, HQ-SAM

Alias: SegmentAnythingin3D, SA3D

Efficient portrait animation with stitching and retargeting control

Sharp Monocular View Synthesis in Less Than a Second

An approach to synthesizing photorealistic views from a single image

PointNet

Deep learning on point sets for 3D classification and segmentation

The Google AI research team explores zero-label learning (training using only synthetic data) in natural language processing and introduces UDG (Unsupervised Data Generation), a training data generation procedure designed to synthesize high-quality training data without human annotation.

3D object detection.

Making JoJo images - Official PyTorch repo for JoJoGAN: One Shot Face Stylization

https://github.com/mchong6/JoJoGAN

A customizable, AI-drawn Japanese anime-style illustration generator

https://waifulabs.com/blog/ai-creativity

Generates images from multiple 3D viewing angles

https://github.com/facebookresearch/StyleNeRF

Generate Amazing Anime Pictures With BigGAN. Just Have Fun

https://github.com/HighCWu/anime_biggan_toy

A lightweight vision library for performing large scale object detection/ instance segmentation.

https://github.com/obss/sahi

Exploiting Diffusion Prior for Real-World Image Super-Resolution

https://iceclear.github.io/projects/stablesr/

Enhance Multi-Camera Tracking Accuracy by Fine-Tuning AI Models with Synthetic Data

https://developer.nvidia.com/blog/enhance-multi-camera-tracking-accuracy-by-fine-tuning-ai-models-with-synthetic-data/

MegaSaM - Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos

https://mega-sam.github.io/

CAT4D: Create Anything in 4D with Multi-View Video Diffusion Models

https://cat-4d.github.io/

Generative Agents - Interactive Simulacra of Human Behavior

https://arxiv.org/abs/2304.03442

(roughly a DOOM screenshot)

Real-time recordings of people playing the game DOOM simulated entirely by the GameNGen neural model.

https://gamengen.github.io/

Tencent's high-resolution 3D asset generation system

https://github.com/Tencent/Hunyuan3D-2

Ultra-fast 3D scene generation model

https://szymanowiczs.github.io/bolt3d