Zero-Shot Object Tracking: we added OpenAI's CLIP as the feature extractor for Deep SORT to create a general-purpose object tracker that can work in conjunction with any object detection model. blog.roboflow.com/zero-shot-obje…

I've been waiting for someone to try this! Image patch level token identities is an interesting way to capture it. Would be interesting to see something on the granularity of image patches, down to pixels - maybe on micro model arxiv.org/abs/2106.08254

en.wikipedia.org/wiki/Extreme_i…

Excited to see our partner Rex at LEDA teaching a class about Roboflow in Taiwan 🇹🇼! ➡ Anyone, anywhere can create a free public Roboflow account and start exploring computer vision. Visit roboflow.com to get started today.

Following up our recent work on zero-shot object tracking - we (Maxwell Stone Brad Dwyer) posted a guide here on how to do object tracking, eliminates labeling your tracks and training another model blog.roboflow.com/object-trackin…

🚨Webinar Inbound: Come check out some really cool technology with Marisa Smith I this Wednesday at 10AM CT. We will be showcasing how you can use Roboflow + Streamlit to deploy computer vision applications in the web browser. Register for free here: lu.ma/mt14sc28

Come to see a cool Streamlit and Roboflow webinar!

👀

An example of using Roboflow Inference API. github.com/hiamitabha/cou… Here is a demo of our Vector being able to recognize another Vector by employing the inference API. youtu.be/YsrOLqVoNJo Very useful resource for robots and other edge devices which lack compute.