正在加载视频...

视频加载失败

返回首页

The Vision Encoder: Our Secret Weapon 🔥 An ultra-accurate vision encoder plate, combined with 5μm-resolution optical feedback, calibrates the H2D’s motion system to achieve consistent 50μm accuracy across the entire build plate. #BambuH2D

BambulabGlobal's profile picture

Bambu Lab

53,651 subscribers

21,379 次观看 • 1 年前 •via X (Twitter)

科学技术#BambuH2D
LIVE

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

This optical device adds accuracy and enables those with poor vision to compete.
0:16

This optical device adds accuracy and enables those with poor vision to compete.

RWB_American

83,873 次观看 • 7 个月前

the encoder
0:11

the encoder

Kat ⊷ the Poet Engineer

103,750 次观看 • 6 个月前

i just ran Google's brand new Unsloth Gemma4 12B dense GGUF on my RTX 4060 using llama.cpp + CUDA 13.2 21 tokens per second. on a budget consumer GPU. locally. no API. no cloud. no subscription. and the benchmarks are absolutely cooked # first let's talk architecture because this is genuinely different every multimodal model you've used has a frozen vision encoder + frozen audio encoder + LLM backbone glued together Gemma 4 12B is different it's a single decoder only transformer. that's it. vision? raw 48×48 pixel patches → one matmul → projected directly into the LLM audio? raw 16kHz signal sliced into 40ms frames → linear projection → same LLM input space no encoder tax. no latency penalty. no fragmented memory to put the encoder savings in perspective: old Gemma 4 26B approach: - 550M param vision encoder (frozen) - 300M param audio encoder (frozen) - LLM backbone Gemma 4 12B: - 35M param vision embedder (a single matmul) - no audio encoder at all - LLM backbone handles EVERYTHING 550M → 35M for vision alone. that's a 15x reduction this is why the gemma-4-12b-it-Q4_K_M.gguf is just 6.6 GBs!!! and it has 256K native context context # Benchmarks: AIME 2026 (math olympiad): 77.5% GPQA Diamond (expert science): 78.8% LiveCodeBench v6 (real code): 72% Codeforces ELO: 1659 MMLU Pro: 77.2% MATH-Vision: 79.7% BigBench Extra Hard: 53% inference → llama.cpp, LM Studio, vLLM, SGLang llamacpp flags: -m "gemma-4-12b-it-Q4_K_M.gguf" -ngl 99 -c 8000 -v --port 8080 Available on huggingface now! Link below
0:18

i just ran Google's brand new Unsloth Gemma4 12B dense GGUF on my RTX 4060 using llama.cpp + CUDA 13.2 21 tokens per second. on a budget consumer GPU. locally. no API. no cloud. no subscription. and the benchmarks are absolutely cooked # first let's talk architecture because this is genuinely different every multimodal model you've used has a frozen vision encoder + frozen audio encoder + LLM backbone glued together Gemma 4 12B is different it's a single decoder only transformer. that's it. vision? raw 48×48 pixel patches → one matmul → projected directly into the LLM audio? raw 16kHz signal sliced into 40ms frames → linear projection → same LLM input space no encoder tax. no latency penalty. no fragmented memory to put the encoder savings in perspective: old Gemma 4 26B approach: - 550M param vision encoder (frozen) - 300M param audio encoder (frozen) - LLM backbone Gemma 4 12B: - 35M param vision embedder (a single matmul) - no audio encoder at all - LLM backbone handles EVERYTHING 550M → 35M for vision alone. that's a 15x reduction this is why the gemma-4-12b-it-Q4_K_M.gguf is just 6.6 GBs!!! and it has 256K native context context # Benchmarks: AIME 2026 (math olympiad): 77.5% GPQA Diamond (expert science): 78.8% LiveCodeBench v6 (real code): 72% Codeforces ELO: 1659 MMLU Pro: 77.2% MATH-Vision: 79.7% BigBench Extra Hard: 53% inference → llama.cpp, LM Studio, vLLM, SGLang llamacpp flags: -m "gemma-4-12b-it-Q4_K_M.gguf" -ngl 99 -c 8000 -v --port 8080 Available on huggingface now! Link below

Alok

277,107 次观看 • 1 个月前

Absolute Encoder Vs Incremental Encoder
0:11

Absolute Encoder Vs Incremental Encoder

S.T.E.M Explorer

622,164 次观看 • 3 个月前

JUAN SOTO WITH AN ABSOLUTE LASER TO GET THE RUNNER AT THE PLATE! 🔥
0:14

JUAN SOTO WITH AN ABSOLUTE LASER TO GET THE RUNNER AT THE PLATE! 🔥

Colin

1,683,899 次观看 • 1 年前

Stepped to the plate with our torpedo Bratt.
0:10

Stepped to the plate with our torpedo Bratt.

New Jersey Devils

75,674 次观看 • 1 年前

Crazy how transformers work! It’s the encoder part
0:18

Crazy how transformers work! It’s the encoder part

Dev Khant

166,618 次观看 • 1 年前

Can a VLM see without a vision encoder? We trained one for $100, inspired by Gemma 4 12B. Latency on an M3 Pro MacBook: 112 ms -> 1.1 ms for the image path 30% lower end-to-end image+LLM The architecture is just: patchify the image -> linear projection with pos embeddings -> LLM Writeup:
0:12

Can a VLM see without a vision encoder? We trained one for $100, inspired by Gemma 4 12B. Latency on an M3 Pro MacBook: 112 ms -> 1.1 ms for the image path 30% lower end-to-end image+LLM The architecture is just: patchify the image -> linear projection with pos embeddings -> LLM Writeup:

Andi Marafioti

59,886 次观看 • 15 天前

🚨 Only on 365Scores: Exclusive Live Match Tracking 🚨 We’ve officially levelled up. Experience football like never before with our Exclusive Live Match Tracker, now powered by industry-leading Computer Vision technology. What makes it a game-changer? 🤖 Computer Vision Precision: Our AI-driven tech tracks every play with pinpoint accuracy. 🏃‍♂️💨 Fluid Motion: See more accurate and realistic player movement across the pitch. ⚡️Enhanced Visuals: A premium, high-tech interpretation of every match moment. 📲 Download the app now to witness the tech in action! 😎 📺 Available currently in UEFA Competitions 🇪🇺, Bundesliga and Bundesliga 2 🇩🇪
0:15

🚨 Only on 365Scores: Exclusive Live Match Tracking 🚨 We’ve officially levelled up. Experience football like never before with our Exclusive Live Match Tracker, now powered by industry-leading Computer Vision technology. What makes it a game-changer? 🤖 Computer Vision Precision: Our AI-driven tech tracks every play with pinpoint accuracy. 🏃‍♂️💨 Fluid Motion: See more accurate and realistic player movement across the pitch. ⚡️Enhanced Visuals: A premium, high-tech interpretation of every match moment. 📲 Download the app now to witness the tech in action! 😎 📺 Available currently in UEFA Competitions 🇪🇺, Bundesliga and Bundesliga 2 🇩🇪

365Scores

94,269 次观看 • 4 个月前

NO LIDAR, NO RADAR – TESLA BETS EVERYTHING ON VISION Tesla’s FSD cameras called the future early – just eyes on the road, silicon style! Vision Facts: •⁠ ⁠GPS + passive optical, nothing else in the stack •⁠ ⁠Road system built for cameras, not sci-fi sensors •⁠ ⁠Superhuman driving is possible with vision alone Tesla called it in 2017 – solve vision, solve autonomy! Source: Elon Musk, TED Talks, 2017
0:18

NO LIDAR, NO RADAR – TESLA BETS EVERYTHING ON VISION Tesla’s FSD cameras called the future early – just eyes on the road, silicon style! Vision Facts: •⁠ ⁠GPS + passive optical, nothing else in the stack •⁠ ⁠Road system built for cameras, not sci-fi sensors •⁠ ⁠Superhuman driving is possible with vision alone Tesla called it in 2017 – solve vision, solve autonomy! Source: Elon Musk, TED Talks, 2017

Mario Nawfal

83,380 次观看 • 9 个月前

Experimenting with Glassmorphism and hand interactions on the Apple Vision Pro. Reflections are the Apple Vision hallmark signature. But how can we achieve refractions? Made with RealityKit
0:13

Experimenting with Glassmorphism and hand interactions on the Apple Vision Pro. Reflections are the Apple Vision hallmark signature. But how can we achieve refractions? Made with RealityKit

Roland Smeenk ᯅ

21,656 次观看 • 1 个月前

The plate has no nutritional value, it's just an empty decorative plate
0:14

The plate has no nutritional value, it's just an empty decorative plate

Didi

1,702,203 次观看 • 6 个月前

Build the team, present the vision, kick ass.
0:11

Build the team, present the vision, kick ass.

Milwaukee Bucks

31,550 次观看 • 11 个月前

I shouldn't be allowed to use Flipnote Encoder
0:19

I shouldn't be allowed to use Flipnote Encoder

Wonder Koopa (Larry Arc)

54,360 次观看 • 6 个月前

The Bambu Lab Vision Plate actually works! Unfortunately, I learned this the hard way after printing 500+ pieces wrong for GatlingTommy MOAB Story Time 😂🤘 #3DPrinting #Nerf
0:18

The Bambu Lab Vision Plate actually works! Unfortunately, I learned this the hard way after printing 500+ pieces wrong for GatlingTommy MOAB Story Time 😂🤘 #3DPrinting #Nerf

Uncle Jessy

15,710 次观看 • 7 个月前

manual control of the monster mwir lens achieved. gonna hook into the magnetic end stop + encoder to replace the proprietary Telic Optics board.
0:13

manual control of the monster mwir lens achieved. gonna hook into the magnetic end stop + encoder to replace the proprietary Telic Optics board.

MacCallister Higgins

21,252 次观看 • 1 年前

Introducing ✨ Aya Vision ✨ - an open-weights model to connect our world through language and vision Aya Vision adds breakthrough multimodal capabilities to our state-of-the-art multilingual 8B and 32B models. 🌿
0:13

Introducing ✨ Aya Vision ✨ - an open-weights model to connect our world through language and vision Aya Vision adds breakthrough multimodal capabilities to our state-of-the-art multilingual 8B and 32B models. 🌿

Cohere Labs

206,570 次观看 • 1 年前

The number plate 🔥🔥🔥
0:11

Sensitive content

The number plate 🔥🔥🔥

The Instigator

74,107 次观看 • 29 天前

Apple Vision Pro supports Direct touch input across the system 🤩 #AppleVisionPro #visionOS #WWDC23 #SpatialComputing #AR #VR #MR
0:17

Apple Vision Pro supports Direct touch input across the system 🤩 #AppleVisionPro #visionOS #WWDC23 #SpatialComputing #AR #VR #MR

Ben Geskin

672,254 次观看 • 3 年前

A teaser of the latest 3D DUSt3R based models we’re presenting at @cvprconference Discover MUSt3R & Pow3R, universal encoder DUNE + research in navigation, vizloc, segmentation & human motion understanding! All our #CVPR2025 papers are here ➡️
0:12

A teaser of the latest 3D DUSt3R based models we’re presenting at @cvprconference Discover MUSt3R & Pow3R, universal encoder DUNE + research in navigation, vizloc, segmentation & human motion understanding! All our #CVPR2025 papers are here ➡️

NAVER LABS Europe

14,222 次观看 • 1 年前