Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

DiffusionGemma-as-Jev (aka djev) running near-real-time vision detection from a mobile phone using its native vision tower. Please don't fall down the stairs!

mmastrac's profile picture

Matt Mastracci

9,393 subscribers

104,402 görüntüleme • 17 gün önce •via X (Twitter)

Bilim & Teknoloji
LIVE

Anya Rossi• Live Now

Private livecam show

15 Yorum

Matt Mastracci profil fotoğrafı
Matt Mastracci17 gün önce

You can even try it yourself! This is the shipping docker container with a packaged server and examples:

Matt Mastracci profil fotoğrafı
Matt Mastracci17 gün önce

Each image is captured from the webcam, tokenized by the model, and the structured query gets answered by DiffusionGemma in about ~300-400ms. This is the stock DiffusionGemma model, no fine tuning, running practically stock vLLM with only a few patches.

Mike Petch profil fotoğrafı
Mike Petch17 gün önce

It didn't do very well identifying the owner of the house (the furry one). Pet????? I'm sure th cat has something to say about that ;-)

Inch profil fotoğrafı
Inch17 gün önce

it could be useful for blind

Richelle🧬 profil fotoğrafı
Richelle🧬17 gün önce

We just shipped open weights jev:

Matt Mastracci profil fotoğrafı
Matt Mastracci17 gün önce

We should chat - would be great to share a single server implementation and make sure we align on multimodal Jev (mmJev?)! Feel free to reach out (DM or email)

Richelle🧬 profil fotoğrafı
Richelle🧬17 gün önce

dm me!

JB profil fotoğrafı
JB17 gün önce

Was setting up a pipeline with YOLO26 + Jev but going to run this and compare.

Chalkers profil fotoğrafı
Chalkers17 gün önce

I’m going to put it in to my robot dog

johnny, INFINITE FUN ENTHUSIAST profil fotoğrafı
johnny, INFINITE FUN ENTHUSIAST17 gün önce

keep going, Matt!

unsever profil fotoğrafı
unsever17 gün önce

Please give it perception of depth and try this again I'll wait

Latent profil fotoğrafı
Latent17 gün önce

I have been following your posts since yesterday. Amazing

AD3 Digital Marketing profil fotoğrafı
AD3 Digital Marketing17 gün önce

Very cool but to big for my 12gb vram Ç_Ç

Simplicity profil fotoğrafı
Simplicity17 gün önce

Do you want T-800s? Because that's how you get T-800s.

ego profil fotoğrafı
ego17 gün önce

I really need to know how much vram jev uses

Benzer Videolar

sharing an easy way to test DJev (Diffusion Gemma Jev), the next evolution of the incredible jev concept and fully open sourced (model, code, api...) imagine jev but it also has: - native image-input support - native image as options!! - real-time video streaming input - as fast as jev (not with images, 600ms) - same price as jev, but can get even cheaper - using Diffusion Gemma adapted - we are seeing on-par performance but happy to have external validation! It is a research preview made for fun, so don't expect stability and the awesome scaling as jev team has delivered! this was a weekend project :) have fun!
1:30

sharing an easy way to test DJev (Diffusion Gemma Jev), the next evolution of the incredible jev concept and fully open sourced (model, code, api...) imagine jev but it also has: - native image-input support - native image as options!! - real-time video streaming input - as fast as jev (not with images, 600ms) - same price as jev, but can get even cheaper - using Diffusion Gemma adapted - we are seeing on-par performance but happy to have external validation! It is a research preview made for fun, so don't expect stability and the awesome scaling as jev team has delivered! this was a weekend project :) have fun!

David Villalón

43,078 görüntüleme • 16 gün önce

Successfully running Ultralytics YOLOv8 on a Jetson Nano with an IMX219 camera. Real-time vision on the edge.
0:30

Successfully running Ultralytics YOLOv8 on a Jetson Nano with an IMX219 camera. Real-time vision on the edge.

Yu Xiang

193,979 görüntüleme • 9 ay önce

Palmer Luckey’s new Eagle Eye helmet blends real time threat detection with enhanced vision tech, giving soldiers a sci fi level view of the battlefield.
1:00

Palmer Luckey’s new Eagle Eye helmet blends real time threat detection with enhanced vision tech, giving soldiers a sci fi level view of the battlefield.

Massimo

57,885 görüntüleme • 10 ay önce

The main event is coming to a phone near you as WWE 2K comes to mobile, exclusively on #NetflixGames later this Fall 2025 #NextOnNetflix
0:33

The main event is coming to a phone near you as WWE 2K comes to mobile, exclusively on #NetflixGames later this Fall 2025 #NextOnNetflix

Netflix Games

706,313 görüntüleme • 1 yıl önce

reality is getting malleable. real girl dancing, the ai copies it perfectly. made this music video using a wan 2.2 animate workflow with native bf16 weights. the stack: • vision: ~200 base images via grok imagine • motion: wan 2.2 animate • pipeline: visual fision beat detection + pose control • music: suno + reaper
2:27

reality is getting malleable. real girl dancing, the ai copies it perfectly. made this music video using a wan 2.2 animate workflow with native bf16 weights. the stack: • vision: ~200 base images via grok imagine • motion: wan 2.2 animate • pipeline: visual fision beat detection + pose control • music: suno + reaper

tetsuo

22,240 görüntüleme • 9 ay önce

Starting today, the BMA is surveilling our borderline day and night using advanced drones flying as high as 7 000 meters at over 40km/h, equipped with night-vision, thermal detection and AI, while body cameras clamp down further on corruption. Digital transformation in action! 🇿🇦
0:27

Starting today, the BMA is surveilling our borderline day and night using advanced drones flying as high as 7 000 meters at over 40km/h, equipped with night-vision, thermal detection and AI, while body cameras clamp down further on corruption. Digital transformation in action! 🇿🇦

Leon Schreiber

263,899 görüntüleme • 1 yıl önce

one of the biggest hurdles in computer vision is that the front end actually matters. you can’t just build models and call it a product. that’s why we’re building Manako. we see it as a “control plane” for visual intelligence. where data, policies, and operators meet. where vision stops being research and starts having real impact in the physical world. this is built for people running real systems. operators who can do 1,000x more when vision is packaged as a real ai product from a dev perspective, we’re already 2 months ahead of plan. now the focus is polish, usability, and connecting it directly into the subnet. build with vision.
5:26

one of the biggest hurdles in computer vision is that the front end actually matters. you can’t just build models and call it a product. that’s why we’re building Manako. we see it as a “control plane” for visual intelligence. where data, policies, and operators meet. where vision stops being research and starts having real impact in the physical world. this is built for people running real systems. operators who can do 1,000x more when vision is packaged as a real ai product from a dev perspective, we’re already 2 months ahead of plan. now the focus is polish, usability, and connecting it directly into the subnet. build with vision.

Max

10,537 görüntüleme • 8 ay önce

i made a computer vision system to monitor the situation - live webcam feed and roboflow RF-DETR for object detection - inputs are fed into SmolVLM which describes what it sees in one sentence - dataviz with vanilla js all running locally in real-time on my macbook air m2
0:27

i made a computer vision system to monitor the situation - live webcam feed and roboflow RF-DETR for object detection - inputs are fed into SmolVLM which describes what it sees in one sentence - dataviz with vanilla js all running locally in real-time on my macbook air m2

AA

72,731 görüntüleme • 7 ay önce

Check out this Circulus Pion robot at Computex 2026. It’s running on a Panther Lake processor, using P-cores for real-time control, E-cores to collect sensor data, the GPU for vision-language-action models, and the NPU for continuous perception.
2:12

Check out this Circulus Pion robot at Computex 2026. It’s running on a Panther Lake processor, using P-cores for real-time control, E-cores to collect sensor data, the GPU for vision-language-action models, and the NPU for continuous perception.

Intel Technology

10,752 görüntüleme • 4 ay önce

UBTech Robotics has showcased its humanoid robot Walker S2 playing tennis with a human. Using real-time vision sensors, it tracks the ball, adjusts its position, and returns shots with good accuracy. The robot can also swap its own batteries, allowing it to operate continuously without human help.
0:33

UBTech Robotics has showcased its humanoid robot Walker S2 playing tennis with a human. Using real-time vision sensors, it tracks the ball, adjusts its position, and returns shots with good accuracy. The robot can also swap its own batteries, allowing it to operate continuously without human help.

Space and Technology

16,121 görüntüleme • 5 ay önce

Check out this guy wearing an Apple Vision using hand-tracking to control a robot in real time. Could this shift more jobs toward teleoperation in the future?
1:23

Check out this guy wearing an Apple Vision using hand-tracking to control a robot in real time. Could this shift more jobs toward teleoperation in the future?

Nathie

32,919 görüntüleme • 2 yıl önce

Humanoid robots can play soccer using a new vision-driven controller, facilitating their victory at real world competitions such as the RoboCup 2025. Learn more in Science #Robotics:
0:36

Humanoid robots can play soccer using a new vision-driven controller, facilitating their victory at real world competitions such as the RoboCup 2025. Learn more in Science #Robotics:

Science Robotics

555,310 görüntüleme • 1 ay önce

Step-3.7-Flash Q4_K_S on DGX Spark (GB10, 128GB): > ~27 tok/s generation > 198B sparse MoE, ~11B active > 256K context, native vision > Agentic / tool-calling / reasoning > Apache 2.0 I added a mobile chat screen on the right showing what 27 tok/s actually feels like streaming on a phone.
0:52

Step-3.7-Flash Q4_K_S on DGX Spark (GB10, 128GB): > ~27 tok/s generation > 198B sparse MoE, ~11B active > 256K context, native vision > Agentic / tool-calling / reasoning > Apache 2.0 I added a mobile chat screen on the right showing what 27 tok/s actually feels like streaming on a phone.

stevibe

19,658 görüntüleme • 4 ay önce

Vision Group shareholder Robert Gumede says their plans after rescuing Tongaat Hulett from liquidation are to diversify the business into an Agri-energy business. He says they don't want to keep it running only as a sugar business. Watch:
2:10

Vision Group shareholder Robert Gumede says their plans after rescuing Tongaat Hulett from liquidation are to diversify the business into an Agri-energy business. He says they don't want to keep it running only as a sugar business. Watch:

Newzroom Afrika

310,608 görüntüleme • 3 ay önce

Why not use the real-world as a game level? This is a real-time augmented reality prototype using a physics based character controller, all running on the web 👀
0:30

Why not use the real-world as a game level? This is a real-time augmented reality prototype using a physics based character controller, all running on the web 👀

Ian Curtis

533,419 görüntüleme • 3 yıl önce

We don’t need a technocrat so much as we need a visionary. We need someone who appreciates that the vision of a country that belongs to its people rather than its corporations is the vision we should hold most dear and protect from all assault. We don’t need more political car mechanics; the problem is that we’re on the wrong road.
1:24

We don’t need a technocrat so much as we need a visionary. We need someone who appreciates that the vision of a country that belongs to its people rather than its corporations is the vision we should hold most dear and protect from all assault. We don’t need more political car mechanics; the problem is that we’re on the wrong road.

Marianne Williamson

14,453 görüntüleme • 2 yıl önce