Weakly Supervised Object Localization via Transformer with Implicit Spatial Calibration

learnable parameter to dynamically adjust the semantic correlations and spatial context intensities for effective information propagation.

Github: https://github.com/164140757/scm

Paper: https://arxiv.org/abs/2207.10447v1

Dataset: https://paperswithcode.com/dataset/cub-200-2011

@computer_science_and_programming

UFO: segmentation 140+ FPS

👉Unified Transformer Framework for Co-Segmentation, Co-Saliency & Salient Object Detection. All in one!


𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Unified framework for co-segmentation
✅Co-segmentation, co-saliency, saliency
✅Block for long-range dependencies
✅Able to reach for 140 FPS in inference
✅The new SOTA on multiple datasets

Paper:
https://arxiv.org/pdf/2203.04708v2.pdf

Code:
https://github.com/suyukun666/UFO


@computer_science_and_programming

Harvard CS109A #DataScience course materials — huge collection free & open!

1. Lecture notes
2. R code, #Python notebooks
3. Lab material
4. Advanced sections
and more ...

https://harvard-iacs.github.io/2019-CS109A/pages/materials.html

@computer_science_and_programming

Resources for performing deep learning on satellite imagery:
- Techniques
- Datasets
- ML best Practice
- Courses
and more ...

@computer_science_and_programming

Omni3D: A Large Benchmark and Model for 3D Object Detection in the Wild

Paper:
https://arxiv.org/pdf/2207.10660.pdf

Github:
https://github.com/facebookresearch/omni3d

Project page:
https://garrickbrazil.com/omni3d/


@computer_science_and_programming

You don't need to spend several $𝟭𝟬𝟬𝟬𝘀 to learn Data Science.❌

Stanford University, Harvard University & Massachusetts Institute of Technology is providing free courses.💥

Here's 8 free Courses that'll teach you better than the paid ones:


1. CS50’s Introduction to Artificial Intelligence with Python (Harvard)

https://lnkd.in/d9CkkfGK

2. Data Science: Machine Learning (Harvard)

https://lnkd.in/dQ7zkCv9

3. Artificial Intelligence (MIT)

https://lnkd.in/dG5BCPen

4. Introduction to Computational Thinking and Data Science (MIT)

https://lnkd.in/ddm5Ckk9

5. Machine Learning (MIT)

https://lnkd.in/dJEjStCw

6. Matrix Methods in Data Analysis, Signal Processing, and Machine Learning (MIT)

https://lnkd.in/dkpyt6qr

7. Statistical Learning (Stanford)

https://lnkd.in/dymn4hbD

8. Mining Massive Data Sets (Stanford)

📍https://lnkd.in/d2uf-FkB

@computer_science_and_programming

SSL4EO-S12: A Large-Scale Multi-Modal, Multi-Temporal Dataset for Self-Supervised Learning in Earth Observation


The dataset consists of unlabeled patch triplets from 251,079 locations across the globe, each patch covering 2640m x 2640m and including 4 seasonal time stamps.

Github:
https://github.com/zhu-xlab/ssl4eo-s12

Paper:
https://arxiv.org/abs/2211.07044v1

Dataset:
https://mediatum.ub.tum.de/1660427

@computer_science_and_programming

Automatically find and fix errors in any ML datasets with cleanlab

This data-centric AI package facilitates machine learning with messy, real-world data by providing clean labels during training.

Github:
https://github.com/cleanlab/cleanlab

Docs:
https://docs.cleanlab.ai/stable/index.html

Examples:
https://github.com/cleanlab/examples

Paper:
https://arxiv.org/abs/2211.13895v1

👉 @computer_science_and_programming

DiffusionInst: Diffusion Model for Instance Segmentation

* DiffusionInst is the first work of diffusion model for instance segmentation

Github:
https://github.com/chenhaoxing/DiffusionInst

Paper:
https://arxiv.org/abs/2212.02773v2

Getting started:
https://github.com/chenhaoxing/DiffusionInst/blob/main/GETTING_STARTED.md

Dataset:
https://paperswithcode.com/dataset/lvis


@computer_science_and_programming

DeepLSD: Line Segment Detection and Refinement with Deep Image Gradients

DeepLSD is a generic line detector that combines the robustness of deep learning with the accuracy of handcrafted detectors. It can be used to extract generic line segments from images in-the-wild, and is suitable for any task requiring high precision, such as homography estimation, visual localization, and 3D reconstruction. By predicting a line distance and angle fields, it can furthermore refine any existing line segments through an optimization


Paper:
https://arxiv.org/abs/2212.07766v1

Github:
https://github.com/cvg/deeplsd

Dataset:
https://paperswithcode.com/dataset/hpatches


@computer_science_and_programming

Accurate and Efficient Stereo Matching via Attention Concatenation Volume

Stereo Depth Estimation

Paper:
https://arxiv.org/pdf/2209.12699.pdf

Github:
https://github.com/gangweiX/Fast-ACVNet

Demo:
https://www.youtube.com/watch?v=az4Z3dp72Zw

ONNX:
ONNX-FastACVNet-Stereo-Depth-Estimation

@computer_science_and_programming

Happy New Year!

Summary of our channel for 2022.
(thanks for curated summary for TGSTAT team)

TGSTAT team: In the new 2023 year, we wish a rapid increase in subscribers, high posts reach, high-quality active audience and, of course, happiness and health.

A traditional present from us is a New Year card with your channel's this year results.

See you in 2023,

@computer_science_and_programming

PACO: Parts and Attributes of Common Objects

Sometimes object detection is not enough and you need more detail about object. Especially, when parts of objects is matters in your task. Then this dataset is for you from Facebook research team.

PACO is a detection dataset that goes beyond traditional object boxes and masks and provides richer annotations such as part masks and attributes. It spans 75 object categories, 456 object-part categories and 55 attributes across image (LVIS) and video (Ego4D) datasets.

Paper:
https://arxiv.org/pdf/2301.01795.pdf

Github:
https://github.com/facebookresearch/paco

Visualization:
https://github.com/facebookresearch/paco/tree/main/notebooks

@computer_science_and_programming

MIT Introduction to Deep Learning - 2023 Starting soon! MIT Intro to DL is one of the most concise AI courses on the web that cover basic deep learning techniques, architectures, and applications.

2023 lectures are starting in just one day, Jan 9th!

Link to register:
http://introtodeeplearning.com

MIT Introduction to Deep Learning The 2022 lectures can be found here:

https://m.youtube.com/playlist?list=PLtBw6njQRU-rwp5__7C0oIVt26ZgjG9NI

👉 @computer_science_and_programming

YOLOv8 is the newest state-of-the-art YOLO model that can be used for object detection, image classification, and instance segmentation tasks. YOLOv8 includes numerous architectural and developer experience changes and improvements over YOLOv5.

Code:
https://github.com/ultralytics/ultralytics

What's New in YOLOv8 ?
https://blog.roboflow.com/whats-new-in-yolov8/

Yolov8 Instance Segmentation (ONNX):
https://github.com/ibaiGorordo/ONNX-YOLOv8-Instance-Segmentation

👉 @computer_science_and_programming

Box2Mask: Box-supervised Instance Segmentation via Level-set Evolution

BoxInstSeg is a toolbox that aims to provide state-of-the-art box-supervised instance segmentation algorithms. It supports instance segmentation with only box annotations.


Github:
https://github.com/LiWentomng/BoxInstSeg

Paper:
https://arxiv.org/pdf/2212.01579.pdf

👉@computer_science_and_programming

GLIGEN: Open-Set Grounded Text-to-Image Generation.

GLIGEN (Grounded-Language-to-Image Generation) a novel approach that builds upon and extends the functionality of existing pre-trained text-to-image diffusion models by enabling them to also be conditioned on grounding inputs.

Project page:
https://gligen.github.io/

Paper:
https://arxiv.org/abs/2301.07093

Github (coming soon):
https://github.com/gligen/GLIGEN

Demo:
https://huggingface.co/spaces/gligen/demo


👉@computer_science_and_programming

Cut and Learn for Unsupervised Object Detection and Instance Segmentation

Cut-and-LEaRn (CutLER) is a simple approach for training object detection and instance segmentation models without human annotations. It outperforms previous SOTA by 2.7 times for AP50 and 2.6 times for AR on 11 benchmarks.

Paper:
https://arxiv.org/pdf/2301.11320.pdf

Github:
https://github.com/facebookresearch/CutLER

Demo:
https://colab.research.google.com/drive/1NgEyFHvOfuA2MZZnfNPWg1w5gSr3HOBb?usp=sharing

👉@computer_science_and_programming

Audio AI Timeline

Here we will keep track of the latest AI models for audio generation, starting in 2023!


▪️SingSong: Generating musical accompaniments from singing
- Paper

▪️AudioLDM: Text-to-Audio Generation with Latent Diffusion Models
- Paper
- Code

▪️Moûsai: Text-to-Music Generation with Long-Context Latent Diffusion
- Paper
- Code

▪️Make-An-Audio: Text-To-Audio Generation with Prompt-Enhanced Diffusion Models
- Paper

▪️Noise2Music

▪️RAVE2
- Paper
- Code

▪️MusicLM: Generating Music From Text
- Paper

▪️Msanii: High Fidelity Music Synthesis on a Shoestring Budget
- Paper
- Code
- HuggingFace

▪️ArchiSound: Audio Generation with Diffusion
- Paper
- Code

▪️VALL-E: Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers
- Paper

👉@computer_science_and_programming