retinify

27 posts

retinify banner
retinify

retinify

@retinify

🚀 Real-Time AI Stereo Vision Library #retinify #robotics

japan Katılım Mayıs 2024
4 Takip Edilen64 Takipçiler
retinify retweetledi
Rerun
Rerun@rerundotio·
Will they score? Follow along with this on-board recording of the @dutchnaoteam during the World Humanoid Robot Games. Rerun is a proud sponsor of the team.
English
0
6
16
990
retinify
retinify@retinify·
@kashu_yamazaki Yes, it is metric depth. This is achieved through stereo calibration, which converts disparity into real-world distances.
English
0
0
1
129
retinify retweetledi
retinify
retinify@retinify·
Real-Time Stereo Matching with retinify! 🚀 Try it on GitHub: github.com/retinify/retin… Features: ✅ Open Source ✅ Works with any stereo camera ✅ Real-time performance on NVIDIA Jetson Orin Nano With retinify, even inexpensive OEM stereo cameras—or a pair of mono cameras—can become high-precision AI stereo cameras! We are visualizing point clouds using @rerundotio ! The C++ API of Rerun is very easy to use, making development much more efficient! #retinify #OpenSource #Robotics #OpenCV #StereoVision #DepthSensing #ROS2
English
1
26
122
12.3K
retinify
retinify@retinify·
@91Oraora We’re not limiting the transparent objects to any specific type. Depending on the use case, we plan to extend training as needed. One practical approach we’re considering is to incorporate a transformer-based refiner for improved handling of transparency.
English
0
0
0
62
retinify
retinify@retinify·
@91Oraora Currently, transparent textures are not supported. However, we believe that with additional training the model could be extended to handle transparent objects, and we’re planning how to approach this in future updates.
English
0
0
1
250
retinify
retinify@retinify·
@Takumi3Dcamera コメントありがとうございます! retinifyはディープラーニングベースのステレオ推定なので、オクルージョン領域についても視差を推定しています。 ただしオクルージョンでは精度が下がりやすいため、必要に応じてLRチェックなどで除外するオプションも用意しています。
日本語
1
2
8
517
Takumi-san@3Dcamera
Takumi-san@3Dcamera@Takumi3Dcamera·
Really !? ステレオにしてはエッジのジャギジャギ感が少ないな。オクルージョンとかどうしてるんだろう?
retinify@retinify

Real-Time Stereo Matching with retinify! 🚀 Try it on GitHub: github.com/retinify/retin… Features: ✅ Open Source ✅ Works with any stereo camera ✅ Real-time performance on NVIDIA Jetson Orin Nano With retinify, even inexpensive OEM stereo cameras—or a pair of mono cameras—can become high-precision AI stereo cameras! We are visualizing point clouds using @rerundotio ! The C++ API of Rerun is very easy to use, making development much more efficient! #retinify #OpenSource #Robotics #OpenCV #StereoVision #DepthSensing #ROS2

日本語
2
3
22
3.4K
retinify retweetledi
avnish sachar
avnish sachar@sacharavnish·
game changing
retinify@retinify

Real-Time Stereo Matching with retinify! 🚀 Try it on GitHub: github.com/retinify/retin… Features: ✅ Open Source ✅ Works with any stereo camera ✅ Real-time performance on NVIDIA Jetson Orin Nano With retinify, even inexpensive OEM stereo cameras—or a pair of mono cameras—can become high-precision AI stereo cameras! We are visualizing point clouds using @rerundotio ! The C++ API of Rerun is very easy to use, making development much more efficient! #retinify #OpenSource #Robotics #OpenCV #StereoVision #DepthSensing #ROS2

English
0
1
2
536
retinify retweetledi
UniflexAI
UniflexAI@UniflexAI·
Here’s why I love the saying: every visual navigation engineer must also be a sensor engineer. Same view, different results. Frame t had great depth—then the next frame? Terrible. Network failure? Nope. Just Realsense’s default 20ms exposure 🤯 — way too high for SLAM.
UniflexAI tweet mediaUniflexAI tweet media
English
0
1
3
515
retinify
retinify@retinify·
@UniflexAI Thank you for trying it out! We find balancing speed and accuracy quite challenging. We’ve measured latency as well — please see the README for details. github.com/retinify/retin…
English
0
0
0
46
UniflexAI
UniflexAI@UniflexAI·
@retinify Hi, we did early test of your model. It's amazing, especially when considering it's parameter size (10MB onnx).
English
1
0
0
75
retinify retweetledi
sasaki@engineer
sasaki@engineer@rsasaki0109·
Retinify is an advanced AI-powered stereo vision library designed for robotics. It enables real-time, high-precision 3D perception by leveraging GPU and NPU acceleration. Its C++ API allows the same code to run seamlessly across various acceleration backends. github.com/retinify/retin…
GIF
English
1
28
132
24K
retinify retweetledi
Shisato Yano
Shisato Yano@4310sy·
Real-Time AI Stereo Vision Library Retinify is an advanced AI-powered stereo vision library designed for robotics. github.com/retinify/retin…
GIF
English
0
11
79
2.6K
retinify
retinify@retinify·
⚠️ This video shows inference results accelerated by NVIDIA TensorRT.
English
0
0
0
232
retinify retweetledi
Rerun
Rerun@rerundotio·
Rerun 0.24 is out! 🥳 This is a big one! Highlights: 🏷️ Tagged components: Log the same component type to the same entity without overwriting 🎥 Video Streams: live stream encoded video to the viewer 😎 Light Mode: It's finally here! 🤖 Built-in URDF Loader: `rr.log_file()` visualizes URDF files
Rerun tweet media
English
1
9
39
6.1K
retinify
retinify@retinify·
Flickering appears in the back left due to the LR consistency check, indicating low disparity reliability.
English
0
0
0
169