Lodestar (@LodestarAI) header
Lodestar (@LodestarAI) profile picture

Lodestar

@LodestarAI

The world’s first real-time active learning #dataannotation platform to accelerate high-quality dataset and #computervision model creation.

 San Jose, California
lodestar.ai
Joined December 2021
  • Tweets37
  • Following11
  • Followers12
7Media
Photo by Lodestar (@LodestarAI)Photo by Lodestar (@LodestarAI)Photo by Lodestar (@LodestarAI)Photo by Lodestar (@LodestarAI)Photo by Lodestar (@LodestarAI)Photo by Lodestar (@LodestarAI)Photo by Lodestar (@LodestarAI)
Happy to share this mini-series where we will be discussing practical insights that lead to success in #computervision projects. Let's start with an intro to #deeplearning and complexity of DL projects. #DataScience bit.ly/3LLiNdB
1
A weekend read #DataScience #deeplearning #AI
Andrej Karpathy (@karpathy) profile pictureAndrej Karpathy
@karpathy31 Mar 2022
“Exploring Plain Vision Transformer Backbones for Object Detection” arxiv.org/abs/2203.16527 Excellent read as usual from the FAIR team. Strong object detection results with only minor tweaks on the vanilla (ViT) Transformer backbone.
Exploring Plain Vision Transformer Backbones for Object Detection

Exploring Plain Vision Transformer Backbones for Object Detection

We explore the plain, non-hierarchical Vision Transformer (ViT) as a backbone network for object detection. This design enables the original ViT architecture to be fine-tuned for object detection...

arxiv.org
1
Done experimenting with a raw equirectangular video and made a quick app to share results. Annotated only 76 frames suggested by active learning, and the model is already doing pretty well. Thoughts? Btw this was done on a .mov file, 1080p, 30fps #deeplearning #AI #DataScience
2
1
A map for your machine learning needs: bit.ly/3lEZkj4
Image from MIT CSAIL (@MIT_CSAIL)
14
215
4
688
In this excellent interview, Ben Thompson (@benthompson) and Jensen Huang CEO of @nvidia dig into the integrated software and hardware approach that has made GPUs the go to hardware for deep learning. #DeepLearning bit.ly/3IKY0Fh
Don't forget to read the latest #AI index report stanford.io/3JTaQm0 by @StanfordHAI @indexing #AIIndex2022 #DataScience
1
Using #computervision and #deeplearning to help detecting rare genetic disorders on early stages is simply awesome . go.nature.com/3NlUQew
1
According to this paper advanced use cases like #deeplearning for video event recognition can still be based on #objectdetection . Agree? bit.ly/3txHVhD #computervision #ai
Some experiments labeling a raw equirectangular video and testing #activelearning #deeplearning on it. Next step is building a test detector... Anyone in the #DataScience community has longer than 1-2 mins videos like this? #AR / #VR are coming soon into our lives :)
Annotating a raw equirectangular video to build a detector using Lodestar.ai platform

ALT Annotating a raw equirectangular video to build a detector using Lodestar.ai platform

4
3
If you are working on object detection / temporal tracking for #AR experiences a good start would be @MetaAI ~3700 hours video dataset #Ego4D . #computervision #datascience #dataset bit.ly/Ego4Ddata
1
3
Interesting real #deeplearning experiments by the @OpenAI team using Transformer on NLP. However, requires a lot of resources → very expensive. openai.com/blog/instruction-…
A very interesting read about interpolation/extrapolation in #deeplearning on high-dimensional datasets by @randall_balestr @an_open_mind and @ylecun #datascience #ai arxiv.org/abs/2110.09485
1
1
Free datasets - Thread 🧵 While working on #computervision projects you probably faced the need for #datasets to use in your #datascience or #machinelearning experiments. Here is a list of 10 free datasets available for download. Feel free to add! Let’s begin with... 👇
Image from Lodestar (@LodestarAI)
10
5
4
10. CIFAR-10 and CIFAR-100 bit.ly/CIFAR-10 CIFAR-10 is a large dataset that consists of 60,000 color images in 10 different classes. CIFAR-100 at the same time has 100 classes containing 600 images each. By @UofTCompSci researchers
9. CelebA bit.ly/CelebA A large-scale face attributes dataset with more than 200K celebrity images, each with 40 attribute annotations, covering large pose variations and background clutter. Has large diversities, large quantities, and rich annotations By @CUHKofficial
8. Cityscapes bit.ly/UrbanStreets The Cityscapes Dataset focuses on semantic understanding of urban street scenes. Numbers: * 50 cities * 5000 images with fine annotations * 2000 images with coarse annotations By @mercedesbenz @maxplanckpress @visinf
7. Open Images V6 bit.ly/OpenImages A dataset of ~9M images annotated with image-level labels, object bounding boxes, object segmentation masks, visual relationships, and localized narratives. By @GoogleAI
6. LabelMe bit.ly/LabelMeMIT An image dataset with annotations. Numbers: * 187,240 images * 62,197 annotated images * 658,992 labeled objects By @MIT_CSAIL
5. YouTube-8M bit.ly/YouTube8M A large-scale labeled video dataset that has millions of @YouTube video IDs, with high-quality machine-generated annotations from a diverse vocabulary of 3,800+ visual entities. Numbers: * 6.1 Million IDs * 3500000 hrs of video By @GoogleAI
4. BDD100K bit.ly/BBD100K A Diverse Driving Dataset for Heterogeneous Multitask Learning. 100K driving videos collected from more than 50K rides. Each video is 40-second long and 30fps. More than 100 million frames in total. By @UCBerkeley @DrFisherYu
1
3. Kinetics bit.ly/Kinetics700 Several large-scale, high-quality #datasets of URL links of up to 650,000 video clips that cover 400/600/700 human action classes, depending on the dataset. Each clip is human annotated with a single action class, ~10 sec. By @DeepMind
Load more