Stars
Serve, optimize and scale PyTorch models in production
BoxMOT: pluggable SOTA tracking modules for segmentation, object detection and pose estimation models
A Trimap-Free Portrait Matting Solution in Real Time [AAAI 2022]
an unofficial screen-scraping "API" for Mint.com
An easy to understand and better performance version of CenterNet
Converts Optical Flow files to images and optionally compiles them to a video. Flow viewer GUI is also available. Check out mockup right from Github Pages:
Code for "Layered Neural Rendering for Retiming People in Video."
Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more
An Unsupervised Learning Framework for Moving Object Detection From Videos
A Simple and Versatile Framework for Object Detection and Instance Recognition
Real-time multi-person tracker using YOLO v3 and deep sort
HOTA (and other) evaluation metrics for Multi-Object Tracking (MOT).
Deformable DETR: Deformable Transformers for End-to-End Object Detection.
FootAndBall: Integrated player and ball detector
Torchreid: Deep learning person re-identification in PyTorch.
Simultaneous object detection and tracking using center points.
[CVPR2021 Oral] End-to-End Video Instance Segmentation with Transformers
Joint detection and tracking model named DEFT, or ``Detection Embeddings for Tracking." Our approach relies on an appearance-based object matching network jointly-learned with an underlying object …
Differentiable convex optimization layers
Differentiable Optimization-Based Modeling for Machine Learning
Run VSCode (codeserver) on Google Colab or Kaggle Notebooks
VISSL is FAIR's library of extensible, modular and scalable components for SOTA Self-Supervised Learning with images.
☁️ Build multimodal AI applications with cloud-native stack
🏄 Scalable embedding, reasoning, ranking for images and sentences with CLIP
SimCLRv2 - Big Self-Supervised Models are Strong Semi-Supervised Learners