It's SLAMSLAM
It's SLAMSLAM
  • 41
  • 18 620
Weekly Spatial AI #-19 DiTer++, RL tutorial, NaVILA, GS-LIVM, Bonsai, MEVIUS, MAN TruckScenes
#ai #slam #robotics #nvidia #diffusion #gaussiansplatting #nvidia
Meeting log: github.com/changh95/WeeklySpatialAI/issues/21
Kakao community: open.kakao.com/o/g8T5kxLb
Facebook: groups/spatialaikr/
0:00:00 - Intro
0:00:30 - Tutorial on reinforcement learning
0:01:15 - NaVILA (Legged Robot Vision-Language-Action Model for Navigation)
0:02:40 - GS-LIVM: Real-Time Photo-Realistic LiDAR-Inertial-Visual Mapping with Gaussian Splatting
0:04:25 - Bonsai C++ library (Octree/Voxel processing library)
0:06:45 - Tesla Optimus walking on rough terrain
0:09:00 - MEVIUS
0:10:40 - DiTer++
0:12:00 - MOANA dataset
0:13:00 - MAN TruckScenes dataset
0:14:40 - Question on DiTer++ (LiDAR interference)
0:16:10 - Question on DiTer++ (How to get GT pose?)
#weeklyspatialai #ai #slam #robotics #korean
มุมมอง: 123

วีดีโอ

Weekly Spatial AI #-18 L3DG, IG-SLAM, ULSR-GS, ROVER, World Labs, MVD2, FastSR-NeRF, Zero-to-Hero
มุมมอง 174หลายเดือนก่อน
#ai #slam #robotics #nvidia #diffusion #gaussiansplatting #nvidia Meeting log: github.com/changh95/WeeklySpatialAI/issues/20 Kakao community: open.kakao.com/o/g8T5kxLb Facebook: groups/spatialaikr/ 0:00:00 - Intro 0:00:40 - L3DG (Latent 3D Gaussian Diffusion) 0:07:20 - IG-SLAM (Instant Gaussian SLAM) 0:08:20 - USLR-GS (Ultra Large-scale Surface Reconstruction Gaussian Splatting wit...
Weekly Spatial AI #-17 MAGiC-SLAM, DROID-Splat, Splat-AD, OVO-SLAM, YOLO-cpp, ALIKED-cpp
มุมมอง 77หลายเดือนก่อน
#ai #slam #robotics #nvidia #diffusion #gaussiansplatting #nvidia Meeting log: github.com/changh95/WeeklySpatialAI/issues/19 Kakao community: open.kakao.com/o/g8T5kxLb Facebook: groups/spatialaikr/ 0:00:00 - MAGiC-SLAM (Multi-agent Gaussian Splatting SLAM) 0:03:00 - DROID-Splat (DROID-SLAM Gaussian Splatting) 0:07:30 - Splat-AD (Image LiDAR rendering for Autonomous Driving) 0:11:05...
[Spatial AI Study] COMO: Compact Mapping and Odometry (Kor)
มุมมอง 39หลายเดือนก่อน
Presenter: Jeesung Kim Date: 2024.12.10 Paper: Dexheimer et al 2024 - COMO: Compact Mapping and Odometry (arxiv.org/abs/2404.03531) Language: Korean
[Spatial AI Study] Learning a Depth Covariance Function (Kor)
มุมมอง 33หลายเดือนก่อน
Presenter: Hyunggi Chang Date: 2024.12.10 Paper: Dexheimer et al 2024 - Learning a Depth Covariance Function (arxiv.org/abs/2303.12157) Language: Korean
[Spatial AI Study] End-to-End Egospheric Spatial Memory (Kor)
มุมมอง 30หลายเดือนก่อน
Presenter: Heokjin Yun Date: 2024.12.10 Paper: Lenton et al 2021 - End-to-End Egospheric Spatial Memory (arxiv.org/abs/2102.07764) Language: Korean
[Spatial AI Study] CodeMapping: Real-Time Dense Mapping for Sparse SLAM using Compact Scene... (Kor)
มุมมอง 34หลายเดือนก่อน
Presenter: Jaemin Lee Date: 2024.11.26 Paper: Matsuki et al 2020 - CodeMapping: Real-Time Dense Mapping for Sparse SLAM using Compact Scene Representations (arxiv.org/abs/2107.08994) Language: Korean
[Spatial AI Study] DeepFactors: Real-Time Probabilistic Dense Monocular SLAM (Kor)
มุมมอง 25หลายเดือนก่อน
Presenter: Heesung Kim Date: 2024.11.26 Paper: Czarnowski et al 2020 - DeepFactors: Real-Time Probabilistic Dense Monocular SLAM (arxiv.org/abs/2001.05049) Language: Korean
[Spatial AI Study] SceneCode: Monocular Dense Semantic Reconstruction using Learned Encoded... (Kor)
มุมมอง 34หลายเดือนก่อน
Presenter: Myungchul Kwak Date: 2024.11.19 Paper: Zhi et al 2019 - SceneCode: Monocular Dense Semantic Reconstruction using Learned Encoded Scene Representations (arxiv.org/abs/1903.06482) Language: Korean
[Spatial AI Study] CodeSLAM - Learning a Compact, Optimisable Representation for Dense VSLAM (Kor)
มุมมอง 25หลายเดือนก่อน
Presenter: Junghyun Park Date: 2024.11.19 Paper: Bloesch et al 2019 - CodeSLAM - Learning a Compact, Optimisable Representation for Dense Visual SLAM (arxiv.org/abs/1804.00874) Language: Korean
[Spatial AI Study] PixRO: Pixel-Distributed Rotational Odometry with GBP (Kor)
มุมมอง 28หลายเดือนก่อน
Presenter: Wonhee Lee Date: 2024.11.12 Paper: Alzugaray et al 2024 - PixRO: Pixel-Distributed Rotational Odometry with Gaussian Belief Propagation (arxiv.org/abs/2406.09726) Language: Korean
[Spatial AI Study] Distributed Simultaneous Localisation and Auto-Calibration using GBP (Kor)
มุมมอง 23หลายเดือนก่อน
Presenter: Jaemin Lee Date: 2024.11.05 Paper: Murai et al 2024 - Distributed Simultaneous Localisation and Auto-Calibration using Gaussian Belief Propagation (arxiv.org/abs/2401.15036) Language: Korean
[Spatial AI Study] Learning in Deep Factor Graphs with Gaussian Belief Propagation (Kor)
มุมมอง 34หลายเดือนก่อน
Presenter: Sunho Kim Date: 2024.11.05 Paper: Nabarro et al 2024 - Learning in Deep Factor Graphs with Gaussian Belief Propagation (arxiv.org/abs/2311.14649) Language: Korean
Tenstorrent QuietBox demo - Running Llama 3.1 70B on vllm
มุมมอง 107หลายเดือนก่อน
Tenstorrent QuietBox demo - Running Llama 3.1 70B on vllm
Weekly Spatial AI #-16 Vision-Language Model (VLM), Large Spatial Model, PLGS, Niantic Scaniverse
มุมมอง 94หลายเดือนก่อน
#ai #slam #robotics #nvidia #diffusion #gaussiansplatting #nvidia Meeting log: github.com/changh95/WeeklySpatialAI/issues/18 Kakao community: open.kakao.com/o/g8T5kxLb Facebook: groups/spatialaikr/ #weeklyspatialai #ai #slam #robotics #korean
Weekly Spatial AI #-15 CoVLA, Fast image processing for VSLAM, 3D object & view generation overview
มุมมอง 57หลายเดือนก่อน
Weekly Spatial AI #-15 CoVLA, Fast image processing for VSLAM, 3D object & view generation overview
Weekly Spatial AI #-14 KISS-Matcher, PyTorch Mobile, Neural fields in robotics, Visual SLAM roadmap
มุมมอง 63หลายเดือนก่อน
Weekly Spatial AI #-14 KISS-Matcher, PyTorch Mobile, Neural fields in robotics, Visual SLAM roadmap
Weekly Spatial AI #-13 Vision-Language Model (VLM), Large Spatial Model, PLGS, Niantic Scaniverse
มุมมอง 142หลายเดือนก่อน
Weekly Spatial AI #-13 Vision-Language Model (VLM), Large Spatial Model, PLGS, Niantic Scaniverse
Weekly Spatial AI #-12 PROSAC, Efficient descriptors, UniTR, DSVT, BEVFusion, MaskBEV
มุมมอง 73หลายเดือนก่อน
Weekly Spatial AI #-12 PROSAC, Efficient descriptors, UniTR, DSVT, BEVFusion, MaskBEV
[Spatial AI Study] A Robot Web for Distributed Many-Device Localisation (Kor)
มุมมอง 1252 หลายเดือนก่อน
[Spatial AI Study] A Robot Web for Distributed Many-Device Localisation (Kor)
[Spatial AI Study] BIT-VO: Visual Odometry at 300 FPS using Binary Features from the Focal Plane (Ko
มุมมอง 952 หลายเดือนก่อน
[Spatial AI Study] BIT-VO: Visual Odometry at 300 FPS using Binary Features from the Focal Plane (Ko
Weekly Spatial AI #-11 CLIP-Clique, FoundPose, CyberCab/Robovan, WildFusion, 𝛼LiDAR, Tenstorrent
มุมมอง 1563 หลายเดือนก่อน
Weekly Spatial AI #-11 CLIP-Clique, FoundPose, CyberCab/Robovan, WildFusion, 𝛼LiDAR, Tenstorrent
Weekly Spatial AI #-10 MonST3R, Depth Pro, EVER, KISS-Matcher, Nano-PGO, ST-P3
มุมมอง 1543 หลายเดือนก่อน
Weekly Spatial AI #-10 MonST3R, Depth Pro, EVER, KISS-Matcher, Nano-PGO, ST-P3
Weekly Spatial AI #9 - MASt3R-SfM, Hyperion, latentSplat, RL meets VO, Aria dataset, NVIDIA GPU
มุมมอง 1393 หลายเดือนก่อน
Weekly Spatial AI #9 - MASt3R-SfM, Hyperion, latentSplat, RL meets VO, Aria dataset, NVIDIA GPU
[Spatial AI Study] Visual odometry using focal-plane sensor-processor (Kor)
มุมมอง 953 หลายเดือนก่อน
[Spatial AI Study] Visual odometry using focal-plane sensor-processor (Kor)
[Spatial AI Study] Incremental Abstraction in Distributed Probabilistic SLAM Graphs (Kor)
มุมมอง 513 หลายเดือนก่อน
[Spatial AI Study] Incremental Abstraction in Distributed Probabilistic SLAM Graphs (Kor)
[Spatial AI Study] A visual introduction to gaussian belief propagation (Kor)
มุมมอง 693 หลายเดือนก่อน
[Spatial AI Study] A visual introduction to gaussian belief propagation (Kor)
[Spatial AI Study] Bundle adjustment on a graph processor (Kor)
มุมมอง 853 หลายเดือนก่อน
[Spatial AI Study] Bundle adjustment on a graph processor (Kor)
[Spatial AI Study] FutureMapping 2: Gaussian Belief Propagation for Spatial AI (Part 1, Kor)
มุมมอง 933 หลายเดือนก่อน
[Spatial AI Study] FutureMapping 2: Gaussian Belief Propagation for Spatial AI (Part 1, Kor)
[Spatial AI study] FutureMapping: The Computational Structure of Spatial AI Systems 논문 리뷰
มุมมอง 953 หลายเดือนก่อน
[Spatial AI study] FutureMapping: The Computational Structure of Spatial AI Systems 논문 리뷰

ความคิดเห็น

  • @andrebarlocher5808
    @andrebarlocher5808 4 วันที่ผ่านมา

    Thank you for this video, really hard to come by reviews and tests for the TT hardware. Also considering getting one for development. Would you say, could the QuietBox handle 100 concurrent requests? 50?

    • @slam_slam_
      @slam_slam_ 4 วันที่ผ่านมา

      @@andrebarlocher5808 not sure. That will depend on how much context window you want to have for each request.

    • @andrebarlocher5808
      @andrebarlocher5808 4 วันที่ผ่านมา

      @@slam_slam_ Let's say, 2k to 3k tokens per request, probably less

  • @ThePoser010
    @ThePoser010 หลายเดือนก่อน

    Was it worth the 15k?

    • @slam_slam_
      @slam_slam_ หลายเดือนก่อน

      @@ThePoser010 I had access to remote server for free 😅 It’s cheaper than 4090 server though, for now.

  • @gftyjhbvtyred
    @gftyjhbvtyred หลายเดือนก่อน

    좋은 자료 잘보고 있습니다 감사합니다. 오늘 진도는 너무 빨라서 천천히 봐야 할거 같습니다 ㅠㅜ

  • @박대성지능로봇전공한
    @박대성지능로봇전공한 หลายเดือนก่อน

    안녕하세요 우선 항상 좋은 영상 올려주셔서 감사합니다. ㅎㅎ 다름이 아니라, Niantic Scaniverse을 깔아 눈 덮인 공터를 스캐닝 해보았는데 잘 안되더라구요. 3D gaussian splatting도 이미지에 특징이 없는 표면을 복원하는 것이 어려운 작업인가요? 혹시 관련하여 아시는 정보가 있으신가 하여 댓글 남겨 봅니다!! 그리고 VLM도 정말 관심 있어하는 주제인데 재밌게 잘 들었습니다!!!! 감사합니다~

    • @slam_slam_
      @slam_slam_ หลายเดือนก่อน

      현우님 답장 전해드립니다! --- 3DGS도 결국 tiles라는 패치 이미지를 기준으로 주변 타일들을 가우시안화해서 만드는 기술인데, 타일의 센트로이드를 못 잡는 경우-예를 들면 눈의 빛 반사나 특징 밀도 부족, 또는 주변 타일들과의 차이점을 모호하게 구분하지 못하는 상황-가우시안을 그릴 수 있는 타일이 튀어버릴 수 있어서, 결국 스캐닝 성능이 떨어질 수밖에 없다고 봅니다. 이건 SLAM에서 로컬 매핑할 때도 자주 나오는 문제 같아요.

  • @slam_slam_
    @slam_slam_ หลายเดือนก่อน

    0:00:00 - Intro 0:01:20 - Meta의 Introduction to Vision-Language Modeling 0:02:40 - VLM 서베이 논문 0:03:20 - Large Spatial Model: End-to-end unposed images to semantic 3D 0:04:45 - Where am I and What will I see? 0:06:05 - PLGS: Robust Panoptic Lifting with 3D Gaussian Splatting 0:07:30 - RANSAC Back to SOTA 0:11:50 - Q&A: VLM이 뭔가요? 0:18:40 - LM-Nav (질문에 답변) 0:27:20 - Q&A: 'RANSAC Back to SOTA가 100% 정확하다고 주장하는 근거는?' 0:34:20 - Niantic Scaniverse 0:36:40 - MVSplat project 0:39:00 - RTGS 0:43:30 - Scaniverse 데모 0:50:10 - Metal API에 대한 잡담 0:52:30 - Niantic 및 미국vs유럽 엔지니어 연봉 관련 잡담

  • @slam_slam_
    @slam_slam_ หลายเดือนก่อน

    0:00:00 - Intro 0:00:40 - Survey on embodied AI 0:03:04 - PROSAC (Old but gold!) 0:10:10 - ORB보다 빠른 feature descriptor 0:12:50 - 박사, 교수, 이직 채용공고 0:17:30 - UniTR 0:26:20 - BEVFusion 0:27:50 - MaskBEV 0:31:45 - Claude 3.5 Sonnet 신기능 0:38:00 - 어쩌다보니 Tenstorrent 이야기

  • @slam_slam_
    @slam_slam_ 3 หลายเดือนก่อน

    #ai #slam #robotics #nvidia #diffusion #gaussiansplatting #nvidia Meeting log: github.com/changh95/WeeklySpatialAI/issues/12 Kakao community: open.kakao.com/o/g8T5kxLb Facebook: facebook.com/groups/spatialaikr/ 0:00:00 - MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion 0:03:35 - Depth Pro: Sharp Monocular Metric Depth in Less Than a Second 0:05:55 - EVER: Exact Volumetric Ellipsoid Rendering for Real-time View Synthesis 0:07:40 - KISS-Matcher: Fast and Robust Point Cloud Registration Revisited 0:10:40 - Diffusion Models in 3D Vision: A Survey 0:14:40 - Nano-PGO 0:19:03 - ST-P3 0:27:20 - [Question] How did you auto-generate graph solver code? 0:34:30 - [Question] Any experiences in developing SLAM with LLM? #weeklyspatialai #ai #slam #robotics #korean

  • @slam_slam_
    @slam_slam_ 3 หลายเดือนก่อน

    0:00:48 - Master-SfM 0:04:20 - Hyperion SLAM - A fast, versatile symbolic Gaussian Belief Propagation framework for Continuous-Time SLAM 0:08:30 - LatentSplat - Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction 0:10:30 - Reinforcement learning meets visual odometry 0:11:30 - Aria digital twin dataset 0:13:50 - PhD intern on Embodied AI at Meta 0:14:20 - [Question] Is COLMAP SfM or just a viewer? 0:17:00 - [Question] Is RL+VO fast enough? 0:21:25 - NVIDIA GPU technology part 2 - Data parallelism 0:25:20 - MPI in NVIDIA GPU 0:30:20 - NCCL in NVIDIA GPU 0:31:30 - GPU profiling 0:34:20 - NeMo for Generative AI

  • @sonny1552
    @sonny1552 3 หลายเดือนก่อน

    너무 좋은 영상 감사드립니다. SLAM 연구에서 전반적인 흐름을 이해하는데 너무 도움이 되는 것 같습니다. 혹시 향후 최신 트랜드(i.e. NeRF, 3D-GS 등)를 반영해서 후속편도 계획이 있으실까요!

    • @slam_slam_
      @slam_slam_ 3 หลายเดือนก่อน

      @@sonny1552 한번 리뉴얼을 할까 고민이 됩니다 ㅎㅎ 새로 업데이트가 되면 알림 드릴게요!

  • @teetanrobotics5363
    @teetanrobotics5363 3 หลายเดือนก่อน

    if possible, could you please make the courses in english ?

    • @slam_slam_
      @slam_slam_ 3 หลายเดือนก่อน

      @@teetanrobotics5363 sure, give me some time and I’ll make a translated version too. Can’t promise on the voice though

  • @grimsk
    @grimsk 4 หลายเดือนก่อน

    회사에 라이카 RTC360 3대, BLK2GO 2대 있는데 반가운 주제네요 :)

  • @leedida2134
    @leedida2134 5 หลายเดือนก่อน

    항상 감사드립니다.

  • @leedida2134
    @leedida2134 5 หลายเดือนก่อน

    항상 감사드립니다!

  • @slam_slam_
    @slam_slam_ 5 หลายเดือนก่อน

    커서가 안보이는걸 방금 알았네요 ㅠㅠ 다음부터는 유의해서 하겠습니다!

  • @leedida2134
    @leedida2134 5 หลายเดือนก่อน

    항상 감사드립니다

  • @hitchhikersguide5758
    @hitchhikersguide5758 5 หลายเดือนก่อน

    스레드 통해 왔음. 좋은 정보 감사.

  • @leedida2134
    @leedida2134 5 หลายเดือนก่อน

    항상 감사드립니다

  • @grimsk
    @grimsk 5 หลายเดือนก่อน

    와 너무 좋아요 :)

  • @박대성지능로봇전공한
    @박대성지능로봇전공한 5 หลายเดือนก่อน

    감사합니다~~ 매주 기대하고 있어용!!

  • @dodeakim
    @dodeakim 6 หลายเดือนก่อน

    감사합니다

  • @slam_slam_
    @slam_slam_ 6 หลายเดือนก่อน

    0:00:00 - FutureMapping: The Computational Structure of Spatial AI Systems 0:14:30 - GLIM: 3D Range-Inertial Localization mapping with GPU-Accelerated Scan Matching Factors 0:24:30 - DeepSLAM: A Robust Monocular SLAM with Unsupervised Deep Learning 0:27:25 - Cross-view Transformers for Real-Time Map-View Semantic Segmentation 0:35:40- Co-RaL: Complementary Radar-Leg Odometry with 4-DoF Optimization and Rolling Contact 0:44:10 - SOLiD: Spatially Organized and Lightweight Global Descriptor for FOV-constrained LiDAR Place Recognition. 0:54:30 - ETPNav: Evolving topological planning for vision-language navigation in continuous environments. 0:57:10 - Learning generalizable feature fields for mobile manipulation. 한글 자막 (cc)를 활성화하시면 자막이 나옵니다!

  • @AllAboutTeslaFSD
    @AllAboutTeslaFSD 2 ปีที่แล้ว

    오옷 최근에는 영상 안올려주시나요?!

    • @slam_slam_
      @slam_slam_ 2 ปีที่แล้ว

      곧 다시 시작해보려고 합니다!

    • @AllAboutTeslaFSD
      @AllAboutTeslaFSD 2 ปีที่แล้ว

      @@slam_slam_ 오오 기대하고 있겠습니다!!

  • @amortalbeing
    @amortalbeing 2 ปีที่แล้ว

    Hi, Is there any chance that you have an english version for this? I really wanted to understand all this but sadly I cant understand a thing except that jusamida thing!

    • @slam_slam_
      @slam_slam_ 2 ปีที่แล้ว

      I’m thinking of rebooting the channel with english contents! :) Subscribe to the channel to follow the updates!

    • @amortalbeing
      @amortalbeing 2 ปีที่แล้ว

      @@slam_slam_ That would be a great idea! I just hope that doesn't take along time ;) looking forward to your English videos

    • @amortalbeing
      @amortalbeing 2 ปีที่แล้ว

      @@slam_slam_ already subbed :) keep up the great job man

  • @공부하는나무-w9z
    @공부하는나무-w9z 3 ปีที่แล้ว

    오픈 톡방 들어갈 수 있을까요? 영상 감사합니다!

    • @slam_slam_
      @slam_slam_ 3 ปีที่แล้ว

      open.kakao.com/o/g8T5kxLb 링크입니다!

    • @yonghyunjo6701
      @yonghyunjo6701 2 ปีที่แล้ว

      안녕하세요. 혹시 오픈챗방 들어가려면 어떤 코드를 입력해야하나요

  • @sejong-ai
    @sejong-ai 4 ปีที่แล้ว

    1:24 목차 (진행방향 요약 : Feature based slam, 종류, limitation, direct slam, 종류, limitation, inertial slam, deep slam) 9:02 옛날 옛적에 수학적으로, 인지적(생물적) 접근* Local Feature 을 찾자. Local Feature 중 가장 괜찮은 Corner 을 잡자. Corner 을 잡았으니 이를 잘 표현할 수 있는 descriptor 을 만들자. 이렇게 만들어진 descriptor 을 잘 matching 해 주자. Feature detector / Feature descriptor / Matching 이 각각 발전해 나갔다. <terminology : Pinhole, Projective geometry*, Geometric feature, edge/line detection (Sobel operator, Hough transform*, Canny Edge detection*> 15:50 Corner point detection <terminology : Harris corner detection and corner ness score*, n point algorithm, RANSAC* > 32:35 Feature detection <terminology : SIFT*, Scale invariance* - pyramid image, Rotation invariance* - image gradient, 50:08 SIFT Feature descriptor*, floating point descriptor, DOG (44:00 Difference of Gaussian) FAST corner detector, FASTER corner, 52:39 BRIEF (Binary Robust Independent Elementry Features, binary descriptor, 1:02:50 ORB*, Oriented FAST, Rotated BRIEF, 1:07:30 AKAZE*, HBST* (Hamming Binary Search Tree)> - 1:16:10 Feature Based SLAM 에서는 PTAM (key frame 에서만 mapping (matching 이 아니고 mapping 임!) 을 한다는 concept) 과 ORB-SLAM 이 Key 패러다임. 최대한 효율적이고 더욱 robust 한 방향으로 발전함. <terminology : 1:16:30 Feature Based SLAM, Mono SLAM, EKF* (Extended Kalman Filter) based, init with known object, 1:22:50 PTAM* (Parallel Tracking and Mapping), manual init, Split tracking and mapping, Keyframe*, 1:29:40 ORB-SLAM*, 1:53:00 REMODE, 1:54:40 Pro-SLAM*> 1:59:40 Feature Based SLAM 의 Limitation - 1:59:40 Direct SLAM <terminology : Direct Photogrammetric method, 2:09:10 DTAM* (Dense tracking and mapping in real-time), feature-less environment, robust against motion-blur, 2:12:30 LSD-SLAM (Large-Scale Direct Monocular SLAM), Semi-dense map, feature bundle, 2:16:05 SVO** (fast Semi Direct Monocular Visual Odometry), 2:20:30 SVO2, 2:26:15 DSO** (direct sparse odometry), sliding window bundle> 2:48:57 Direct SLAM Limitation - 2:48:57 Visual inertial SLAM <terminology : 3:02:12 IMU Bias, diverge 3:04:06 trajectory, extrinsic calibration, VI initialization, 3:09:55 MSCKF* (multi-scale constraint kalman filter), pre-integration, iSAM*, GTSAM*>

  • @sejong-ai
    @sejong-ai 4 ปีที่แล้ว

    27:00 목차 overview 37:00 발표자 소개 38:30 slam 배경지식 (정의, 예시, 특징 및 요구조건, 유사기술들 path planning, visual odom, loop closure, structure from motion, mapping, photogrammetry, image stitching ) 1:02:40 slam 의 종류 1:32:00 deep slam 1:38:30 컴퓨터 및 장비 추천 2:12:55 GPU를 사용하거나 DEEP LEARNING BASED SLAM, Visual slam 의 종류 (feature based = sparse / direct based = dense / deep learning based slam 의 차이를 중심으로) 3:37:07 pahse 2 - digital image low level 요약 3:37:21 biological vision vs silicon based vision - polarisation camera - 4:28:30 event camera - 4:33:00 rgb camera 고르는법 (sensor resolution) - 4:54:30 camera lens 에 대해서 (spatial resolution : 다들을때까지 적기

  • @서정훈-m4c
    @서정훈-m4c 4 ปีที่แล้ว

    27:00 강의시작

  • @skulim1466
    @skulim1466 4 ปีที่แล้ว

    Any chance to have this lecture in English? Looks like a great summary of Visual SLAM and state-of-the-art algorithms.

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      Thanks for the interest! I'll see what I can do.

    • @shengkaiwang2722
      @shengkaiwang2722 4 ปีที่แล้ว

      ​@@slam_slam_ Really hope to see a version presented in English or that of having English subtitles. Thanks!

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      @@shengkaiwang2722 Hi, thanks for your interest in this talk :) Unfortunately, I'm finding it difficult to make an English version or having English subtitles. Instead, I've made a visual-slam study roadmap on my github - I tried to cover as much as possible, so until I deliver a talk in English in the future I hope this roadmap will help you study SLAM. github.com/changh95/visual-slam-roadmap

  • @endland1
    @endland1 4 ปีที่แล้ว

    잘 생겼네요.. 똑똑하고, 사기 캐릭턴데...

  • @HongLab
    @HongLab 4 ปีที่แล้ว

    감사합니다.

  • @raptor-1905
    @raptor-1905 4 ปีที่แล้ว

    블로그가 안들어가지는데 주소가 바뀌었나요?

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      이슈가 조금 있는 것 같네요. cv-learn.com 로 접속해주시면 됩니다

  • @chpyeah
    @chpyeah 4 ปีที่แล้ว

    안녕하세요? 좋은 영상 감사드립니다. slamkr에서도 많은 도움 받고 있습니다. 혹시 본 영상의 슬라이드는 어디에 업로드되어있나요?

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      안녕하세요! 슬라이드는 현재 업로드 되어있진 않습니다. 추후 보기 편하시게 분할 업로드는 예정되어 있습니다 :)

  • @kimchi_taco
    @kimchi_taco 4 ปีที่แล้ว

    새 아이패드 카메라 모듈에 라이다도 달려나오는데, 어떻게 생각하시나 궁금하네요. 모터달려서 돌아가는 라이다가 아니라 그냥 모바일 카메라 만한 라이다입니다. 개인적으로 SLAM센서의 인더스트리 스탠다드가 될것 같다는 느낌인데요. 잘 모르는 입장에서 앞으로 어떻게 될지 예상은 한다면, RGB카메라: DeepVO식으로 완전 e2e deep learning으로 visual odometry. IMU: FPS가 느린 카메라의 RT를 보정해주는 서포터 역할 포인트 Lidar: 십여개 포인트의 뎁스 ground truth를 줌으로서, 기존 SLAM의 약점이었던 scaleless한 문제와, 프레임 마다 RT 에러가 쌓이는 문제를 잡는 서포터 역할 제 생각은 아마 테슬라도 결국 이렇게 RGB카메라+포인트라이다 로 가지 않을까 합니다. 라이다 모터 돌아가는게 극혐이었는데, 안그래도 되니. 작고 싼것 같구요. 이 방식 인더스트리표준 될 것 같다는 느낌입니다. 한편 뎁스카메라는 노이즈가 많고, sparse하고, 햇볓이 강하거나 거리가 멀면 안되니, 그냥 카메라써서 딥러닝 하는거에 대체되면서 사라질것 같구요.

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      좋은 의견 공유 감사합니다 :) 저는 이번 아이패드 발표에 대해 조금 다른 시각으로 보고 있는데요, 1. 양산되는 모바일 기기에 탑재될 수 있는 lidar 기술의 발전과 양산 인프라 구축 2. lidar 데이터를 안정적으로 실시간으로 돌릴 수 있을만큼의 모바일 프로세서 발전 3. 애플 글래스와 함께 애플 eco-system의 다양한 분야에서 증강현실을 적용하려는 움직임 이라고 봅니다. 애플의 솔리드 스테이트 라이다 탑재가 자율주행의 라이다 시스템에 얼마나 영향을 미칠지에 대해서는 잘 모르겠습니다 ㅎㅎ...

  • @kimchi_taco
    @kimchi_taco 4 ปีที่แล้ว

    다른분야에서 일하지만, SLAM을 알고 싶은 엔지니어에게 너무도 좋은 강의였습니다. 헬리콥터를 타고 지형을 개괄적으로 확인하는것 같은 기분이었습니다. 떠먹여 주셔서 감사! 3:13:00 왜 ARCore가 잔디밭 같은데서 멍청이 되는지 알았네요 :)

  • @jjy0923
    @jjy0923 4 ปีที่แล้ว

    감사합니다 형기님!

  • @kimchi_taco
    @kimchi_taco 4 ปีที่แล้ว

    멋진 강의 감사합니다. 스피치쪽 ML엔지니어인데요, 유다시티 self-driving 나노디그리에서 SLAM을 접하고 좀더 알고 싶었는데, 많은 궁금점이 풀렸습니다. cv-learn.com도 즐겁게 읽겠습니다! 한편으로는 스피치 쪽에서도 2-3년 전만해도 딥러닝보다 트레디셔널한 HMM방법이 더 좋다라는 느낌이었는데, 지금은 완전 e2e 딥러닝으로 대체되었습니다. 아마 SLAM도 같은 길을 가게될것 같습니다. 2-3년후엔 다 e2e 딥슬램을 GPU나 edge accelerator에서 돌리는 방식으로요.

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      안녕하세요! 반갑습니다 :) 저도 딥러닝 슬램은 이제 시작이라고 보고 있습니다! SLAM 쪽에서도 ImageNet 같은 데이터셋이 생긴다면 폭발적으로 연구가 진행될 수 있다고 생각합니다. 재밌는 연구 팔로우 하면서 종종 cv-learn 블로그에 올리겠습니다 ㅎㅎ 감사합니다!

  • @slam_slam_
    @slam_slam_ 4 ปีที่แล้ว

    @0:00 - 소개 @8:38 - Local Features @1:16:09 - Feature-based SLAM @1:59:37 - Direct SLAM @2:44:40 - Visual-Inertial SLAM @3:58:10 - SLAM with Deep Learning

  • @dohyungkwon3544
    @dohyungkwon3544 4 ปีที่แล้ว

    앞부분 2시간이 짤린거같아요 ㅠㅠ 좋은 자료 감사합니다!

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      유투브에서 긴 라이브를 하면 자동으로 영상 편집되면서 그런 것 같습니다! 시간이 지나면 다시 돌아오더라구요 ㅎㅎ 양해부탁드립니다!

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      ㅠㅠ 지금 다시 보니 유투브쪽에서 프로세싱 버그가 있어서 업로드가 안되고 있네요. 구글링해보니 꽤 유명한 버그인 것 같은데, 시간이 해결해줄 것 같습니다. 일단 유투브에 문의 넣어놨습니다. 알려주셔서 감사합니다! EDIT: 해결되었습니다!

    • @dohyungkwon3544
      @dohyungkwon3544 4 ปีที่แล้ว

      @@slam_slam_ 감사합니다! 어제 라이브로 보다가 일이 있어서 다 못봐서요! 좋은 자료 감사합니다!

  • @willpark1579
    @willpark1579 4 ปีที่แล้ว

    매번 감사합니다 😁

  • @최소영-e8t
    @최소영-e8t 4 ปีที่แล้ว

    강의 정말 감사합니다~ 혹시 본 강의의 ppt 자료는 공개 안하시나욤??

    • @slam_slam_
      @slam_slam_ 4 ปีที่แล้ว

      허가를 받고 온 자료들이 있기 때문에 PPT 공개는 어려울 것 같습니다! :)

  • @slam_slam_
    @slam_slam_ 4 ปีที่แล้ว

    @41:00 ~ @1:15:43 Visual-SLAM이란? + SLAM 의 종류 @1:15:44 ~ @1:40:25 Visual-SLAM 공부하기 꿀팁 + 로드맵 @1:40:25 ~ @2:15:08 CV / SLAM 장비 추천 @2:15:10 ~ @2:27:08 GPU를 사용하는 SLAM / Deep-SLAM에 대한 전망 @2:27:09 ~ @2:37:40 Visual-SLAM 기술의 최앞단은 증강현실이다?? (지극히 개인적인 의견) @3:06:06 ~ @3:39:46 제대로 카메라 / 디지털 이미지에 대해 이해하기! @3:39:46 ~ @3:54:49 Silicon-based vision 이해하기 @3:55:50 ~ @4:35:20 다양한 카메라 / 이미징 기법 알아보기 @4:35:25 ~ @5:19:20 RGB 카메라/렌즈 고르기 + 구매하는 방법 @5:19:20 ~ @5:30:51 머신비전/RGB-D 카메라 데모 + SLAM을 위한 카메라 셋팅법

  • @touchtoo4816
    @touchtoo4816 4 ปีที่แล้ว

    이런 무료 꿀강의 라이브로 못봐서 아쉽네요 ㅠㅠ 녹화본 공개해주셔서 감사합니다!

  •  4 ปีที่แล้ว

    Quite good video, awesome! Would you like to be TH-cam friends? :]

  • @happydrawing7309
    @happydrawing7309 4 ปีที่แล้ว

    도박사이트 인줄...

  • @slam_slam_
    @slam_slam_ 4 ปีที่แล้ว

    [코로나19 사태 진정을 간절히 바라며...😌😌 Visual SLAM의 기초에 관한 전액기부 기술 공유 세미나!] 코로나 사태의 최전선에 계시는 분들을 위해 후원금 전액을 전국재해구호협회에 기부 예정입니다. 투네이션 후원 링크: toon.at/donate/vslam_covid19 관련 커뮤니티: SLAM 기술 오픈 카카오톡 채팅방 : open.kakao.com/o/g8T5kxLb 페이스북 SLAM 커뮤니티 : facebook.com/groups/slamkr/ Computer Vision / Visual-SLAM 블로그: www.cv-learn.com 함께 하시는 모든 분들께 감사드립니다 :) @41:00 ~ @1:15:43 Visual-SLAM이란? + SLAM 의 종류 @1:15:44 ~ @1:40:25 Visual-SLAM 공부하기 꿀팁 + 로드맵 @1:40:25 ~ @2:15:08 CV / SLAM 장비 추천 @2:15:10 ~ @2:27:08 GPU를 사용하는 SLAM / Deep-SLAM에 대한 전망 @2:27:09 ~ @2:37:40 Visual-SLAM 기술의 최앞단은 증강현실이다?? (지극히 개인적인 의견) @3:06:06 ~ @3:39:46 제대로 카메라 / 디지털 이미지에 대해 이해하기! @3:39:46 ~ @3:54:49 Silicon-based vision 이해하기 @3:55:50 ~ @4:35:20 다양한 카메라 / 이미징 기법 알아보기 @4:35:25 ~ @5:19:20 RGB 카메라/렌즈 고르기 + 구매하는 방법 @5:19:20 ~ @5:30:51 머신비전/RGB-D 카메라 데모 + SLAM을 위한 카메라 셋팅법