SlideShare ist ein Scribd-Unternehmen logo
1 von 46
Downloaden Sie, um offline zu lesen
Unsupervised Collaborative Learning of
Keyframe Detection and Visual Odometry
Towards Monocular Deep SLAM [Sheng & Xu+, ICCV’19]
東京大学 相澤研究室
M2 金子 真也
1
本論文
• Unsupervised Collaborative Learning of Keyframe Detection
and Visual Odometry Towards Monocular Deep SLAM
– 著者: L. Sheng, D. Xu, W. Ouyang and X. Wang
– 所属: Beihang University, Oxford, SenseTime
– 採択会議: ICCV2019
2
本論文
• Unsupervised Collaborative Learning of Keyframe Detection
and Visual Odometry Towards Monocular Deep SLAM
– 著者: L. Sheng, D. Xu, W. Ouyang and X. Wang
– 所属: Beihang University, Oxford, SenseTime
– 採択会議: ICCV2019
– Monocular Deep SLAMを実現したいという強い気持ちの論文
– わかりみが深い
3
Introduction
• Visual SLAM
– 3D reconstruction + Camera pose estimation
– 両者の同時最適化 (Bundle Adjustment)
Direct Sparse Odometry [Engel+, TPAMI’18]
4
Introduction
• Deep Learning for Visual SLAM (Deep SLAM)
End-to-end Deep SLAMDL helps SLAM
SfMLearner [Zhou+, CVPR’17]
CNN-SLAM [Tateno+, CVPR’17]
CodeSLAM [Tateno+, CVPR’18]
DeepTAM [Zhou+, ECCV’18]
This figure is from Tombari’s presentation slide @ ICCVW.
5
Introduction
• Deep Learning for Visual SLAM (Deep SLAM)
End-to-end Deep SLAMDL helps SLAM
CNN-SLAM [Tateno+, CVPR’17]
CodeSLAM [Tateno+, CVPR’18]
DeepTAM [Zhou+, ECCV’18]
SfMLearner [Zhou+, CVPR’17]
本論文の目標は,
この領域での最高のDeep SLAMを作ること
This figure is from Tombari’s presentation slide @ ICCVW.
6
Related works
• SfMLearner [Zhou+, CVPR’17]
– 古典的なSfMを応用し, UnsupervisedにDeep SLAMを実現
– Training
• Photometric errorを最小化するように学習
7
Related works
• SfMLearner [Zhou+, CVPR’17]
– 古典的なSfMを応用し, UnsupervisedにDeep SLAMを実現
– Inference
• 入力画像の奥行き画像と, 2視点間のカメラ姿勢をCNNで回帰
8
Related works
• SfMLearner [Zhou+, CVPR’17]
– 古典的なSfMを応用し, UnsupervisedにDeep SLAMを実現
– Inference
• 入力画像の奥行き画像と, 2視点間のカメラ姿勢をCNNで回帰
より従来のVSLAMに近い
Deep SLAMを実現するためには???
9
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
10
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
画像 𝑍𝑍𝑗𝑗 画像 𝑍𝑍𝑗𝑗+1
𝒖𝒖𝑖𝑖,𝑗𝑗+1特徴点 𝒖𝒖𝑖𝑖,𝑗𝑗
カメラ姿勢 [𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗]
𝒖𝒖𝑖𝑖,𝑗𝑗
投影点
𝑓𝑓 𝐗𝐗𝑖𝑖 𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗) Bundle
𝑍𝑍𝑗𝑗+1
[𝐑𝐑𝑗𝑗+1, 𝐭𝐭𝑗𝑗+1]
𝒖𝒖𝑖𝑖,𝑗𝑗+1
画像 𝑍𝑍𝑗𝑗
3D位置 𝐗𝐗𝑖𝑖
11
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
画像 𝑍𝑍𝑗𝑗 画像 𝑍𝑍𝑗𝑗+1
𝒖𝒖𝑖𝑖,𝑗𝑗+1特徴点 𝒖𝒖𝑖𝑖,𝑗𝑗
最適化
画像 𝑍𝑍𝑗𝑗
𝒖𝒖𝑖𝑖,𝑗𝑗
Bundle
[𝐑𝐑𝑗𝑗+1, 𝐭𝐭𝑗𝑗+1]
𝑍𝑍𝑗𝑗+1
𝒖𝒖𝑖𝑖,𝑗𝑗+1
3D位置 𝐗𝐗𝑖𝑖
カメラ姿勢 [𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗]
投影点
𝑓𝑓 𝐗𝐗𝑖𝑖 𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗)
12
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
• 三次元地図に, 三次元点とその点が属するカメラ画像を登録
• カメラ画像をKeyframe (KF)と呼ぶ
画像 𝑍𝑍𝑗𝑗 画像 𝑍𝑍𝑗𝑗+1
𝒖𝒖𝑖𝑖,𝑗𝑗+1特徴点 𝒖𝒖𝑖𝑖,𝑗𝑗
最適化
画像 𝑍𝑍𝑗𝑗
𝒖𝒖𝑖𝑖,𝑗𝑗
Bundle
[𝐑𝐑𝑗𝑗+1, 𝐭𝐭𝑗𝑗+1]
𝑍𝑍𝑗𝑗+1
𝒖𝒖𝑖𝑖,𝑗𝑗+1
3D位置 𝐗𝐗𝑖𝑖
カメラ姿勢 [𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗]
投影点
𝑓𝑓 𝐗𝐗𝑖𝑖 𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗)
Keyframe
13
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
• 三次元地図に, 三次元点とその点が属するカメラ画像を登録
• カメラ画像をKeyframe (KF)と呼ぶ
Keyframe
[1] ORB-SLAM2 for Monocular, Stereo and RGB-D Cameras [Mur-Artal+, ToR17]
14
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
• 三次元地図に, 三次元点とその点が属するカメラ画像を登録
• カメラ画像をKeyframe (KF)と呼ぶ
– Keyframeの選び方
1. 重複を避けるようにある程度間隔を空ける
15
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
• 三次元地図に, 三次元点とその点が属するカメラ画像を登録
• カメラ画像をKeyframe (KF)と呼ぶ
– Keyframeの選び方
1. 重複を避けるようにある程度間隔を空ける
2. 十分な地図が作れなくなるのでそれなりに必要
16
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
• 三次元地図に, 三次元点とその点が属するカメラ画像を登録
• カメラ画像をKeyframe (KF)と呼ぶ
– Keyframeの選び方
1. 重複を避けるようにある程度間隔を空ける
2. 十分な地図が作れなくなるのでそれなりに必要
→ 職人技のような挿入条件の設定が必要
17
Related works
• VSLAM
– VSLAMで最も重要な要素はBundle Adjustment (BA)
• 三次元地図に, 三次元点とその点が属するカメラ画像を登録
• カメラ画像をKeyframe (KF)と呼ぶ
– Keyframeの選び方
1. 重複を避けるようにある程度間隔を空ける
2. 十分な地図が作れなくなるのでそれなりに必要
→ 職人技のような挿入条件の設定が必要
– この選択をCNNで実現し, SfMLearnerに組み込めないか?
18
Proposed method
• SfMLearner with KF selection
– KF選択を行いながら, 三次元復元とカメラ姿勢推定を行うような
Deep SLAMの実現
19
Proposed method
• SfMLearner with KF selection
– Unsupervised collaborative learning
• 三次元復元 + カメラ姿勢推定 + KF選択 ←New!!!
KF selection network
Depth + Camera pose network
(Visual Odometry)
Depth + Camera pose network
(Visual Odometry)
20
Proposed method
• SfMLearner with KF selection
– Unsupervised collaborative learning
• 三次元復元 + カメラ姿勢推定 + KF選択 ←New!!!
KF selection network
- 2枚の画像間のsimilarity
scoreを回帰
- このscoreに応じてKFの
選択を行う
21
Proposed method
• SfMLearner with KF selection
– Unsupervised collaborative learning
• 三次元復元 + カメラ姿勢推定 + KF選択 ←New!!!
Depth + Camera pose network
(Visual Odometry) KF selection network
22
Proposed method
• Training
– Data pair
• Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1}
• Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Nearest Keyframe
2nd nearest Keyframe
23
Proposed method
• Training
– Data pair
• Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1}
• Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
– Visual Odometry (≈SfMLearner)
• Photometric error + cycle consistency + depth smooth term
Target 𝐈𝐈𝑡𝑡
𝐃𝐃𝑡𝑡
Reference 𝐈𝐈𝑟𝑟
𝐃𝐃𝑟𝑟
Warped ref 𝐈𝐈𝑡𝑡←𝑟𝑟
𝐃𝐃𝑡𝑡
𝐃𝐃𝑟𝑟
Photometric error Cycle Consistency
t
r
Target 𝐈𝐈𝑡𝑡Warped tgt 𝐈𝐈𝑟𝑟←𝑡𝑡
Warped2
tgt 𝐈𝐈𝑡𝑡←𝑟𝑟←𝑡𝑡
Warped tgt 𝐈𝐈𝑟𝑟←𝑡𝑡
24
Proposed method
• Training
– Data pair
• Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1}
• Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
– Visual Odometry (≈SfMLearner)
• Photometric error + cycle consistency + depth smooth term
– Keyframe selection
• Triplet loss
<𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑝𝑝>
s
t
p
0.1
n
大 小
25
Proposed method
• Training
– Data pair
• Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1}
• Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
– Visual Odometry (≈SfMLearner)
• Photometric error + cycle consistency + depth smooth term
– Keyframe selection
• Triplet loss
<𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑝𝑝>
<𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑛𝑛>
s
t
p
0.1
n
大 小
大 小0.8
26
Proposed method
• Training
– Data pair
• Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1}
• Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
– Visual Odometry (≈SfMLearner)
• Photometric error + cycle consistency + depth smooth term
– Keyframe selection
• Triplet loss
<𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑝𝑝>
<𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑛𝑛>
s
t
p
0.1
n
大 小
大 小0.8
27
Proposed method
• Training
– Data pair
• Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1}
• Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
– Visual Odometry (≈SfMLearner)
• Photometric error + cycle consistency + depth smooth term
– Keyframe selection
• Triplet loss
KFはどのように選ばれるのか?
28
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(I𝑝𝑝, I𝑡𝑡) > th:
Insert tgt frame I𝑡𝑡 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
Model
29
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {𝓘𝓘𝒔𝒔, 𝐈𝐈𝒑𝒑, 𝐈𝐈𝒏𝒏}
Train all the model
if iteration > 200 &
Similarity(I𝑝𝑝, I𝑡𝑡) > th:
Insert tgt frame I𝑡𝑡 as KF
Merge KF
ℐ𝑠𝑠
{𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
𝐈𝐈𝑡𝑡
Model
KF pool 𝒫𝒫 𝐾𝐾
Dataset
30
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, I𝑝𝑝, I𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(I𝑝𝑝, I𝑡𝑡) > th:
Insert tgt frame I𝑡𝑡 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
ℐ𝑠𝑠
{𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
𝐈𝐈𝑡𝑡
Model Loss
Train
31
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, I𝑝𝑝, I𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(I𝑝𝑝, I𝑡𝑡) > th:
Insert tgt frame I𝑡𝑡 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
ℐ𝑠𝑠
{𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
𝐈𝐈𝑡𝑡
Model Loss
Train
32
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th:
Insert tgt frame 𝐈𝐈𝒕𝒕 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
ℐ𝑠𝑠
𝐈𝐈𝑝𝑝
𝐈𝐈𝑡𝑡
Model Score
33
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th:
Insert tgt frame 𝐈𝐈𝒕𝒕 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
ℐ𝑠𝑠
𝐈𝐈𝑝𝑝
𝐈𝐈𝑡𝑡
Model Score
34
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th:
Insert tgt frame 𝐈𝐈𝒕𝒕 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
ℐ𝑠𝑠
𝐈𝐈𝑝𝑝
𝐈𝐈𝑡𝑡
Model Score
35
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th:
Insert tgt frame 𝐈𝐈𝒕𝒕 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
ℐ𝑠𝑠
𝐈𝐈𝑝𝑝
𝐈𝐈𝑡𝑡
Model Score
36
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(I𝑝𝑝, I𝑡𝑡) > th:
Insert tgt frame I𝑡𝑡 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
𝐈𝐈𝑛𝑛𝐈𝐈𝑝𝑝
Model
ℐ𝑠𝑠
𝐈𝐈𝑡𝑡
Scores
37
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(I𝑝𝑝, I𝑡𝑡) > th:
Insert tgt frame I𝑡𝑡 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
Model
38
Proposed method
• Training
– Keyframe update & management
Random KF initialization
for epoch:
for iteration:
Choose training pair {ℐ𝑠𝑠, I𝑝𝑝, I𝑛𝑛}
Train all the model
if iteration > 200 &
Similarity(𝐼𝐼𝑝𝑝, 𝐼𝐼𝑡𝑡) > th:
Insert tgt frame 𝐼𝐼𝑡𝑡 as KF
Merge KF
KF pool 𝒫𝒫 𝐾𝐾
Dataset
Model
この操作を繰り返すことで
KF poolの最適化を行う
39
Experimental results
• KITTI dataset
– Monocular Depth Estimation
KF selectionによって学習データを調整することで, 学習が安定し
推定精度も高くなる
40
Experimental results
• KITTI dataset
– Monocular Depth Estimation
KF selectionによって学習データを調整することで, 学習が安定し
推定精度も高くなる
41
Experimental results
• KITTI dataset
– Absolute Trajectory Error (ATE)
KF selectionがdata augmentationの効果を持ち, 結果としてカメラ
姿勢の推定精度が向上
42
Experimental results
• KITTI dataset
– Average Rotation Errors
とはいえカメラの回転の推定精度はORB-SLAM[Mur-Artal, TOR15]には
勝てていない状況
43
Experimental results
• KITTI dataset
– Keyframe selection
• カメラが並進する場所では, 均一になるように選択
• カメラが回転する場所では, 変化が激しいのでより刻んだ選択
44
Experimental results
• KITTI dataset
– Ablation study
Depth推定
カメラ軌跡推定
45
Conclusion
• SfMLearner with KF selection
– VSLAMで最も重要なKF selectionを, SfMLearnerの枠組みに追加
– UnsupervisedでKF selectionを学習する手法を提案
– 従来手法よりも高精度な奥行き推定, カメラ姿勢推定を達成.
• 感想
– 従来人手の緻密な設計が必要だったKF selectionを, unsupervisedに
CNNで学習し実現した点が新しく非常に面白い
– KF selectionだけでなく, Bundle Adjustment等の最適化要素も追加
できるとDeep SLAMの実現により近付きそう

Weitere ähnliche Inhalte

Was ist angesagt?

Visual-SLAM技術を利用した 果樹園の3次元圃場地図の作成
Visual-SLAM技術を利用した果樹園の3次元圃場地図の作成Visual-SLAM技術を利用した果樹園の3次元圃場地図の作成
Visual-SLAM技術を利用した 果樹園の3次元圃場地図の作成Masahiro Tsukano
 
ORB-SLAMの手法解説
ORB-SLAMの手法解説ORB-SLAMの手法解説
ORB-SLAMの手法解説Masaya Kaneko
 
MIRU2013チュートリアル:SIFTとそれ以降のアプローチ
MIRU2013チュートリアル:SIFTとそれ以降のアプローチMIRU2013チュートリアル:SIFTとそれ以降のアプローチ
MIRU2013チュートリアル:SIFTとそれ以降のアプローチHironobu Fujiyoshi
 
関東コンピュータビジョン勉強会
関東コンピュータビジョン勉強会関東コンピュータビジョン勉強会
関東コンピュータビジョン勉強会nonane
 
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~SSII
 
20180527 ORB SLAM Code Reading
20180527 ORB SLAM Code Reading20180527 ORB SLAM Code Reading
20180527 ORB SLAM Code ReadingTakuya Minagawa
 
確率モデルを用いた3D点群レジストレーション
確率モデルを用いた3D点群レジストレーション確率モデルを用いた3D点群レジストレーション
確率モデルを用いた3D点群レジストレーションKenta Tanaka
 
SLAMチュートリアル大会資料(ORB-SLAM)
SLAMチュートリアル大会資料(ORB-SLAM)SLAMチュートリアル大会資料(ORB-SLAM)
SLAMチュートリアル大会資料(ORB-SLAM)Masaya Kaneko
 
3次元レジストレーション(PCLデモとコード付き)
3次元レジストレーション(PCLデモとコード付き)3次元レジストレーション(PCLデモとコード付き)
3次元レジストレーション(PCLデモとコード付き)Toru Tamaki
 
SSII2018TS: 3D物体検出とロボットビジョンへの応用
SSII2018TS: 3D物体検出とロボットビジョンへの応用SSII2018TS: 3D物体検出とロボットビジョンへの応用
SSII2018TS: 3D物体検出とロボットビジョンへの応用SSII
 
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~SSII
 
LiDAR点群と画像とのマッピング
LiDAR点群と画像とのマッピングLiDAR点群と画像とのマッピング
LiDAR点群と画像とのマッピングTakuya Minagawa
 
Direct Sparse Odometryの解説
Direct Sparse Odometryの解説Direct Sparse Odometryの解説
Direct Sparse Odometryの解説Masaya Kaneko
 
画像認識の初歩、SIFT,SURF特徴量
画像認識の初歩、SIFT,SURF特徴量画像認識の初歩、SIFT,SURF特徴量
画像認識の初歩、SIFT,SURF特徴量takaya imai
 
CNN-SLAMざっくり
CNN-SLAMざっくりCNN-SLAMざっくり
CNN-SLAMざっくりEndoYuuki
 
SLAM開発における課題と対策の一例の紹介
SLAM開発における課題と対策の一例の紹介SLAM開発における課題と対策の一例の紹介
SLAM開発における課題と対策の一例の紹介miyanegi
 
2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentation
2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentation2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentation
2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentationTakuya Minagawa
 
オープンソース SLAM の分類
オープンソース SLAM の分類オープンソース SLAM の分類
オープンソース SLAM の分類Yoshitaka HARA
 

Was ist angesagt? (20)

Structure from Motion
Structure from MotionStructure from Motion
Structure from Motion
 
Visual-SLAM技術を利用した 果樹園の3次元圃場地図の作成
Visual-SLAM技術を利用した果樹園の3次元圃場地図の作成Visual-SLAM技術を利用した果樹園の3次元圃場地図の作成
Visual-SLAM技術を利用した 果樹園の3次元圃場地図の作成
 
ORB-SLAMの手法解説
ORB-SLAMの手法解説ORB-SLAMの手法解説
ORB-SLAMの手法解説
 
MIRU2013チュートリアル:SIFTとそれ以降のアプローチ
MIRU2013チュートリアル:SIFTとそれ以降のアプローチMIRU2013チュートリアル:SIFTとそれ以降のアプローチ
MIRU2013チュートリアル:SIFTとそれ以降のアプローチ
 
関東コンピュータビジョン勉強会
関東コンピュータビジョン勉強会関東コンピュータビジョン勉強会
関東コンピュータビジョン勉強会
 
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
 
20180527 ORB SLAM Code Reading
20180527 ORB SLAM Code Reading20180527 ORB SLAM Code Reading
20180527 ORB SLAM Code Reading
 
確率モデルを用いた3D点群レジストレーション
確率モデルを用いた3D点群レジストレーション確率モデルを用いた3D点群レジストレーション
確率モデルを用いた3D点群レジストレーション
 
SLAMチュートリアル大会資料(ORB-SLAM)
SLAMチュートリアル大会資料(ORB-SLAM)SLAMチュートリアル大会資料(ORB-SLAM)
SLAMチュートリアル大会資料(ORB-SLAM)
 
3次元レジストレーション(PCLデモとコード付き)
3次元レジストレーション(PCLデモとコード付き)3次元レジストレーション(PCLデモとコード付き)
3次元レジストレーション(PCLデモとコード付き)
 
SSII2018TS: 3D物体検出とロボットビジョンへの応用
SSII2018TS: 3D物体検出とロボットビジョンへの応用SSII2018TS: 3D物体検出とロボットビジョンへの応用
SSII2018TS: 3D物体検出とロボットビジョンへの応用
 
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
SSII2019TS: 実践カメラキャリブレーション ~カメラを用いた実世界計測の基礎と応用~
 
LiDAR点群と画像とのマッピング
LiDAR点群と画像とのマッピングLiDAR点群と画像とのマッピング
LiDAR点群と画像とのマッピング
 
Direct Sparse Odometryの解説
Direct Sparse Odometryの解説Direct Sparse Odometryの解説
Direct Sparse Odometryの解説
 
画像認識の初歩、SIFT,SURF特徴量
画像認識の初歩、SIFT,SURF特徴量画像認識の初歩、SIFT,SURF特徴量
画像認識の初歩、SIFT,SURF特徴量
 
20180424 orb slam
20180424 orb slam20180424 orb slam
20180424 orb slam
 
CNN-SLAMざっくり
CNN-SLAMざっくりCNN-SLAMざっくり
CNN-SLAMざっくり
 
SLAM開発における課題と対策の一例の紹介
SLAM開発における課題と対策の一例の紹介SLAM開発における課題と対策の一例の紹介
SLAM開発における課題と対策の一例の紹介
 
2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentation
2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentation2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentation
2018/12/28 LiDARで取得した道路上点群に対するsemantic segmentation
 
オープンソース SLAM の分類
オープンソース SLAM の分類オープンソース SLAM の分類
オープンソース SLAM の分類
 

Ähnlich wie Unsupervised Collaborative Learning of Keyframe Detection and Visual Odometry Towards Monocular Deep SLAMの解説

Review: Incremental Few-shot Instance Segmentation [CDM]
Review: Incremental Few-shot Instance Segmentation [CDM]Review: Incremental Few-shot Instance Segmentation [CDM]
Review: Incremental Few-shot Instance Segmentation [CDM]Dongmin Choi
 
Explaining the decisions of image/video classifiers
Explaining the decisions of image/video classifiersExplaining the decisions of image/video classifiers
Explaining the decisions of image/video classifiersVasileiosMezaris
 
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object Detector
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object DetectorPR-270: PP-YOLO: An Effective and Efficient Implementation of Object Detector
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object DetectorJinwon Lee
 
Realtime pothole detection system using improved CNN Models
Realtime pothole detection system using improved CNN ModelsRealtime pothole detection system using improved CNN Models
Realtime pothole detection system using improved CNN Modelsnithinsai2992
 
[CIKM 2014] Deviation-Based Contextual SLIM Recommenders
[CIKM 2014] Deviation-Based Contextual SLIM Recommenders[CIKM 2014] Deviation-Based Contextual SLIM Recommenders
[CIKM 2014] Deviation-Based Contextual SLIM RecommendersYONG ZHENG
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsScott Clark
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsSigOpt
 
Deep Learning Introduction - WeCloudData
Deep Learning Introduction - WeCloudDataDeep Learning Introduction - WeCloudData
Deep Learning Introduction - WeCloudDataWeCloudData
 
Weight watcher Bay Area ACM Feb 28, 2022
Weight watcher Bay Area ACM Feb 28, 2022 Weight watcher Bay Area ACM Feb 28, 2022
Weight watcher Bay Area ACM Feb 28, 2022 Charles Martin
 
Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D...
 Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D... Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D...
Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D...Databricks
 
Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...
Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...
Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...Joo-Haeng Lee
 
Decision Forests and discriminant analysis
Decision Forests and discriminant analysisDecision Forests and discriminant analysis
Decision Forests and discriminant analysispotaters
 
Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...
Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...
Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...Ansys
 
Cross-validation aggregation for forecasting
Cross-validation aggregation for forecastingCross-validation aggregation for forecasting
Cross-validation aggregation for forecastingDevon Barrow
 
PR-232: AutoML-Zero:Evolving Machine Learning Algorithms From Scratch
PR-232:  AutoML-Zero:Evolving Machine Learning Algorithms From ScratchPR-232:  AutoML-Zero:Evolving Machine Learning Algorithms From Scratch
PR-232: AutoML-Zero:Evolving Machine Learning Algorithms From ScratchSunghoon Joo
 
Willump: Optimizing Feature Computation in ML Inference
Willump: Optimizing Feature Computation in ML InferenceWillump: Optimizing Feature Computation in ML Inference
Willump: Optimizing Feature Computation in ML InferenceDatabricks
 
Object Tracking with Instance Matching and Online Learning
Object Tracking with Instance Matching and Online LearningObject Tracking with Instance Matching and Online Learning
Object Tracking with Instance Matching and Online LearningJui-Hsin (Larry) Lai
 
Weakly-Supervised Sound Event Detection with Self-Attention
Weakly-Supervised Sound Event Detection with Self-AttentionWeakly-Supervised Sound Event Detection with Self-Attention
Weakly-Supervised Sound Event Detection with Self-AttentionNU_I_TODALAB
 

Ähnlich wie Unsupervised Collaborative Learning of Keyframe Detection and Visual Odometry Towards Monocular Deep SLAMの解説 (20)

Review: Incremental Few-shot Instance Segmentation [CDM]
Review: Incremental Few-shot Instance Segmentation [CDM]Review: Incremental Few-shot Instance Segmentation [CDM]
Review: Incremental Few-shot Instance Segmentation [CDM]
 
Explaining the decisions of image/video classifiers
Explaining the decisions of image/video classifiersExplaining the decisions of image/video classifiers
Explaining the decisions of image/video classifiers
 
Temporal Segment Network
Temporal Segment NetworkTemporal Segment Network
Temporal Segment Network
 
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object Detector
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object DetectorPR-270: PP-YOLO: An Effective and Efficient Implementation of Object Detector
PR-270: PP-YOLO: An Effective and Efficient Implementation of Object Detector
 
Realtime pothole detection system using improved CNN Models
Realtime pothole detection system using improved CNN ModelsRealtime pothole detection system using improved CNN Models
Realtime pothole detection system using improved CNN Models
 
[CIKM 2014] Deviation-Based Contextual SLIM Recommenders
[CIKM 2014] Deviation-Based Contextual SLIM Recommenders[CIKM 2014] Deviation-Based Contextual SLIM Recommenders
[CIKM 2014] Deviation-Based Contextual SLIM Recommenders
 
autoTVM
autoTVMautoTVM
autoTVM
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning Models
 
Using Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning ModelsUsing Bayesian Optimization to Tune Machine Learning Models
Using Bayesian Optimization to Tune Machine Learning Models
 
Deep Learning Introduction - WeCloudData
Deep Learning Introduction - WeCloudDataDeep Learning Introduction - WeCloudData
Deep Learning Introduction - WeCloudData
 
Weight watcher Bay Area ACM Feb 28, 2022
Weight watcher Bay Area ACM Feb 28, 2022 Weight watcher Bay Area ACM Feb 28, 2022
Weight watcher Bay Area ACM Feb 28, 2022
 
Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D...
 Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D... Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D...
Apache Spark Based Hyper-Parameter Selection and Adaptive Model Tuning for D...
 
Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...
Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...
Calibration Issues in FRC: Camera, Projector, Kinematics based Hybrid Approac...
 
Decision Forests and discriminant analysis
Decision Forests and discriminant analysisDecision Forests and discriminant analysis
Decision Forests and discriminant analysis
 
Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...
Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...
Use of FEA to Improve the Design of Suspension Springs for Reciprocating Comp...
 
Cross-validation aggregation for forecasting
Cross-validation aggregation for forecastingCross-validation aggregation for forecasting
Cross-validation aggregation for forecasting
 
PR-232: AutoML-Zero:Evolving Machine Learning Algorithms From Scratch
PR-232:  AutoML-Zero:Evolving Machine Learning Algorithms From ScratchPR-232:  AutoML-Zero:Evolving Machine Learning Algorithms From Scratch
PR-232: AutoML-Zero:Evolving Machine Learning Algorithms From Scratch
 
Willump: Optimizing Feature Computation in ML Inference
Willump: Optimizing Feature Computation in ML InferenceWillump: Optimizing Feature Computation in ML Inference
Willump: Optimizing Feature Computation in ML Inference
 
Object Tracking with Instance Matching and Online Learning
Object Tracking with Instance Matching and Online LearningObject Tracking with Instance Matching and Online Learning
Object Tracking with Instance Matching and Online Learning
 
Weakly-Supervised Sound Event Detection with Self-Attention
Weakly-Supervised Sound Event Detection with Self-AttentionWeakly-Supervised Sound Event Detection with Self-Attention
Weakly-Supervised Sound Event Detection with Self-Attention
 

Mehr von Masaya Kaneko

GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説
GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説
GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説Masaya Kaneko
 
論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])
論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])
論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])Masaya Kaneko
 
Neural scene representation and rendering の解説(第3回3D勉強会@関東)
Neural scene representation and rendering の解説(第3回3D勉強会@関東)Neural scene representation and rendering の解説(第3回3D勉強会@関東)
Neural scene representation and rendering の解説(第3回3D勉強会@関東)Masaya Kaneko
 
論文読み会2018 (CodeSLAM)
論文読み会2018 (CodeSLAM)論文読み会2018 (CodeSLAM)
論文読み会2018 (CodeSLAM)Masaya Kaneko
 
Dynamic Routing Between Capsules
Dynamic Routing Between CapsulesDynamic Routing Between Capsules
Dynamic Routing Between CapsulesMasaya Kaneko
 
論文読み会(DeMoN;CVPR2017)
論文読み会(DeMoN;CVPR2017)論文読み会(DeMoN;CVPR2017)
論文読み会(DeMoN;CVPR2017)Masaya Kaneko
 
コンピュータ先端ガイド2巻3章勉強会(SVM)
コンピュータ先端ガイド2巻3章勉強会(SVM)コンピュータ先端ガイド2巻3章勉強会(SVM)
コンピュータ先端ガイド2巻3章勉強会(SVM)Masaya Kaneko
 

Mehr von Masaya Kaneko (7)

GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説
GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説
GN-Net: The Gauss-Newton Loss for Deep Direct SLAMの解説
 
論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])
論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])
論文読み会@AIST (Deep Virtual Stereo Odometry [ECCV2018])
 
Neural scene representation and rendering の解説(第3回3D勉強会@関東)
Neural scene representation and rendering の解説(第3回3D勉強会@関東)Neural scene representation and rendering の解説(第3回3D勉強会@関東)
Neural scene representation and rendering の解説(第3回3D勉強会@関東)
 
論文読み会2018 (CodeSLAM)
論文読み会2018 (CodeSLAM)論文読み会2018 (CodeSLAM)
論文読み会2018 (CodeSLAM)
 
Dynamic Routing Between Capsules
Dynamic Routing Between CapsulesDynamic Routing Between Capsules
Dynamic Routing Between Capsules
 
論文読み会(DeMoN;CVPR2017)
論文読み会(DeMoN;CVPR2017)論文読み会(DeMoN;CVPR2017)
論文読み会(DeMoN;CVPR2017)
 
コンピュータ先端ガイド2巻3章勉強会(SVM)
コンピュータ先端ガイド2巻3章勉強会(SVM)コンピュータ先端ガイド2巻3章勉強会(SVM)
コンピュータ先端ガイド2巻3章勉強会(SVM)
 

Kürzlich hochgeladen

CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Servicegiselly40
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...gurkirankumar98700
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024Scott Keck-Warren
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slidevu2urc
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...
Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...
Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...shyamraj55
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfEnterprise Knowledge
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersThousandEyes
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure servicePooja Nehwal
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsEnterprise Knowledge
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationRidwan Fadjar
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Allon Mureinik
 
Google AI Hackathon: LLM based Evaluator for RAG
Google AI Hackathon: LLM based Evaluator for RAGGoogle AI Hackathon: LLM based Evaluator for RAG
Google AI Hackathon: LLM based Evaluator for RAGSujit Pal
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking MenDelhi Call girls
 

Kürzlich hochgeladen (20)

CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Service
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...
Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...
Automating Business Process via MuleSoft Composer | Bangalore MuleSoft Meetup...
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
My Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 PresentationMy Hashitalk Indonesia April 2024 Presentation
My Hashitalk Indonesia April 2024 Presentation
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)
 
Google AI Hackathon: LLM based Evaluator for RAG
Google AI Hackathon: LLM based Evaluator for RAGGoogle AI Hackathon: LLM based Evaluator for RAG
Google AI Hackathon: LLM based Evaluator for RAG
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men
 

Unsupervised Collaborative Learning of Keyframe Detection and Visual Odometry Towards Monocular Deep SLAMの解説

  • 1. Unsupervised Collaborative Learning of Keyframe Detection and Visual Odometry Towards Monocular Deep SLAM [Sheng & Xu+, ICCV’19] 東京大学 相澤研究室 M2 金子 真也
  • 2. 1 本論文 • Unsupervised Collaborative Learning of Keyframe Detection and Visual Odometry Towards Monocular Deep SLAM – 著者: L. Sheng, D. Xu, W. Ouyang and X. Wang – 所属: Beihang University, Oxford, SenseTime – 採択会議: ICCV2019
  • 3. 2 本論文 • Unsupervised Collaborative Learning of Keyframe Detection and Visual Odometry Towards Monocular Deep SLAM – 著者: L. Sheng, D. Xu, W. Ouyang and X. Wang – 所属: Beihang University, Oxford, SenseTime – 採択会議: ICCV2019 – Monocular Deep SLAMを実現したいという強い気持ちの論文 – わかりみが深い
  • 4. 3 Introduction • Visual SLAM – 3D reconstruction + Camera pose estimation – 両者の同時最適化 (Bundle Adjustment) Direct Sparse Odometry [Engel+, TPAMI’18]
  • 5. 4 Introduction • Deep Learning for Visual SLAM (Deep SLAM) End-to-end Deep SLAMDL helps SLAM SfMLearner [Zhou+, CVPR’17] CNN-SLAM [Tateno+, CVPR’17] CodeSLAM [Tateno+, CVPR’18] DeepTAM [Zhou+, ECCV’18] This figure is from Tombari’s presentation slide @ ICCVW.
  • 6. 5 Introduction • Deep Learning for Visual SLAM (Deep SLAM) End-to-end Deep SLAMDL helps SLAM CNN-SLAM [Tateno+, CVPR’17] CodeSLAM [Tateno+, CVPR’18] DeepTAM [Zhou+, ECCV’18] SfMLearner [Zhou+, CVPR’17] 本論文の目標は, この領域での最高のDeep SLAMを作ること This figure is from Tombari’s presentation slide @ ICCVW.
  • 7. 6 Related works • SfMLearner [Zhou+, CVPR’17] – 古典的なSfMを応用し, UnsupervisedにDeep SLAMを実現 – Training • Photometric errorを最小化するように学習
  • 8. 7 Related works • SfMLearner [Zhou+, CVPR’17] – 古典的なSfMを応用し, UnsupervisedにDeep SLAMを実現 – Inference • 入力画像の奥行き画像と, 2視点間のカメラ姿勢をCNNで回帰
  • 9. 8 Related works • SfMLearner [Zhou+, CVPR’17] – 古典的なSfMを応用し, UnsupervisedにDeep SLAMを実現 – Inference • 入力画像の奥行き画像と, 2視点間のカメラ姿勢をCNNで回帰 より従来のVSLAMに近い Deep SLAMを実現するためには???
  • 10. 9 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA)
  • 11. 10 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) 画像 𝑍𝑍𝑗𝑗 画像 𝑍𝑍𝑗𝑗+1 𝒖𝒖𝑖𝑖,𝑗𝑗+1特徴点 𝒖𝒖𝑖𝑖,𝑗𝑗 カメラ姿勢 [𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗] 𝒖𝒖𝑖𝑖,𝑗𝑗 投影点 𝑓𝑓 𝐗𝐗𝑖𝑖 𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗) Bundle 𝑍𝑍𝑗𝑗+1 [𝐑𝐑𝑗𝑗+1, 𝐭𝐭𝑗𝑗+1] 𝒖𝒖𝑖𝑖,𝑗𝑗+1 画像 𝑍𝑍𝑗𝑗 3D位置 𝐗𝐗𝑖𝑖
  • 12. 11 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) 画像 𝑍𝑍𝑗𝑗 画像 𝑍𝑍𝑗𝑗+1 𝒖𝒖𝑖𝑖,𝑗𝑗+1特徴点 𝒖𝒖𝑖𝑖,𝑗𝑗 最適化 画像 𝑍𝑍𝑗𝑗 𝒖𝒖𝑖𝑖,𝑗𝑗 Bundle [𝐑𝐑𝑗𝑗+1, 𝐭𝐭𝑗𝑗+1] 𝑍𝑍𝑗𝑗+1 𝒖𝒖𝑖𝑖,𝑗𝑗+1 3D位置 𝐗𝐗𝑖𝑖 カメラ姿勢 [𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗] 投影点 𝑓𝑓 𝐗𝐗𝑖𝑖 𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗)
  • 13. 12 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) • 三次元地図に, 三次元点とその点が属するカメラ画像を登録 • カメラ画像をKeyframe (KF)と呼ぶ 画像 𝑍𝑍𝑗𝑗 画像 𝑍𝑍𝑗𝑗+1 𝒖𝒖𝑖𝑖,𝑗𝑗+1特徴点 𝒖𝒖𝑖𝑖,𝑗𝑗 最適化 画像 𝑍𝑍𝑗𝑗 𝒖𝒖𝑖𝑖,𝑗𝑗 Bundle [𝐑𝐑𝑗𝑗+1, 𝐭𝐭𝑗𝑗+1] 𝑍𝑍𝑗𝑗+1 𝒖𝒖𝑖𝑖,𝑗𝑗+1 3D位置 𝐗𝐗𝑖𝑖 カメラ姿勢 [𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗] 投影点 𝑓𝑓 𝐗𝐗𝑖𝑖 𝐑𝐑𝑗𝑗, 𝐭𝐭𝑗𝑗) Keyframe
  • 14. 13 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) • 三次元地図に, 三次元点とその点が属するカメラ画像を登録 • カメラ画像をKeyframe (KF)と呼ぶ Keyframe [1] ORB-SLAM2 for Monocular, Stereo and RGB-D Cameras [Mur-Artal+, ToR17]
  • 15. 14 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) • 三次元地図に, 三次元点とその点が属するカメラ画像を登録 • カメラ画像をKeyframe (KF)と呼ぶ – Keyframeの選び方 1. 重複を避けるようにある程度間隔を空ける
  • 16. 15 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) • 三次元地図に, 三次元点とその点が属するカメラ画像を登録 • カメラ画像をKeyframe (KF)と呼ぶ – Keyframeの選び方 1. 重複を避けるようにある程度間隔を空ける 2. 十分な地図が作れなくなるのでそれなりに必要
  • 17. 16 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) • 三次元地図に, 三次元点とその点が属するカメラ画像を登録 • カメラ画像をKeyframe (KF)と呼ぶ – Keyframeの選び方 1. 重複を避けるようにある程度間隔を空ける 2. 十分な地図が作れなくなるのでそれなりに必要 → 職人技のような挿入条件の設定が必要
  • 18. 17 Related works • VSLAM – VSLAMで最も重要な要素はBundle Adjustment (BA) • 三次元地図に, 三次元点とその点が属するカメラ画像を登録 • カメラ画像をKeyframe (KF)と呼ぶ – Keyframeの選び方 1. 重複を避けるようにある程度間隔を空ける 2. 十分な地図が作れなくなるのでそれなりに必要 → 職人技のような挿入条件の設定が必要 – この選択をCNNで実現し, SfMLearnerに組み込めないか?
  • 19. 18 Proposed method • SfMLearner with KF selection – KF選択を行いながら, 三次元復元とカメラ姿勢推定を行うような Deep SLAMの実現
  • 20. 19 Proposed method • SfMLearner with KF selection – Unsupervised collaborative learning • 三次元復元 + カメラ姿勢推定 + KF選択 ←New!!! KF selection network Depth + Camera pose network (Visual Odometry)
  • 21. Depth + Camera pose network (Visual Odometry) 20 Proposed method • SfMLearner with KF selection – Unsupervised collaborative learning • 三次元復元 + カメラ姿勢推定 + KF選択 ←New!!! KF selection network - 2枚の画像間のsimilarity scoreを回帰 - このscoreに応じてKFの 選択を行う
  • 22. 21 Proposed method • SfMLearner with KF selection – Unsupervised collaborative learning • 三次元復元 + カメラ姿勢推定 + KF選択 ←New!!! Depth + Camera pose network (Visual Odometry) KF selection network
  • 23. 22 Proposed method • Training – Data pair • Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1} • Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Nearest Keyframe 2nd nearest Keyframe
  • 24. 23 Proposed method • Training – Data pair • Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1} • Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} – Visual Odometry (≈SfMLearner) • Photometric error + cycle consistency + depth smooth term Target 𝐈𝐈𝑡𝑡 𝐃𝐃𝑡𝑡 Reference 𝐈𝐈𝑟𝑟 𝐃𝐃𝑟𝑟 Warped ref 𝐈𝐈𝑡𝑡←𝑟𝑟 𝐃𝐃𝑡𝑡 𝐃𝐃𝑟𝑟 Photometric error Cycle Consistency t r Target 𝐈𝐈𝑡𝑡Warped tgt 𝐈𝐈𝑟𝑟←𝑡𝑡 Warped2 tgt 𝐈𝐈𝑡𝑡←𝑟𝑟←𝑡𝑡 Warped tgt 𝐈𝐈𝑟𝑟←𝑡𝑡
  • 25. 24 Proposed method • Training – Data pair • Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1} • Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} – Visual Odometry (≈SfMLearner) • Photometric error + cycle consistency + depth smooth term – Keyframe selection • Triplet loss <𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑝𝑝> s t p 0.1 n 大 小
  • 26. 25 Proposed method • Training – Data pair • Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1} • Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} – Visual Odometry (≈SfMLearner) • Photometric error + cycle consistency + depth smooth term – Keyframe selection • Triplet loss <𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑝𝑝> <𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑛𝑛> s t p 0.1 n 大 小 大 小0.8
  • 27. 26 Proposed method • Training – Data pair • Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1} • Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} – Visual Odometry (≈SfMLearner) • Photometric error + cycle consistency + depth smooth term – Keyframe selection • Triplet loss <𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑝𝑝> <𝐈𝐈𝑡𝑡, 𝐈𝐈𝑠𝑠, 𝐈𝐈𝑛𝑛> s t p 0.1 n 大 小 大 小0.8
  • 28. 27 Proposed method • Training – Data pair • Sequential frames ℐ𝑠𝑠 ={𝐈𝐈𝑡𝑡−1, 𝐈𝐈𝑡𝑡, 𝐈𝐈𝑡𝑡+1} • Keyframes {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} – Visual Odometry (≈SfMLearner) • Photometric error + cycle consistency + depth smooth term – Keyframe selection • Triplet loss KFはどのように選ばれるのか?
  • 29. 28 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Train all the model if iteration > 200 & Similarity(I𝑝𝑝, I𝑡𝑡) > th: Insert tgt frame I𝑡𝑡 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset Model
  • 30. 29 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {𝓘𝓘𝒔𝒔, 𝐈𝐈𝒑𝒑, 𝐈𝐈𝒏𝒏} Train all the model if iteration > 200 & Similarity(I𝑝𝑝, I𝑡𝑡) > th: Insert tgt frame I𝑡𝑡 as KF Merge KF ℐ𝑠𝑠 {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} 𝐈𝐈𝑡𝑡 Model KF pool 𝒫𝒫 𝐾𝐾 Dataset
  • 31. 30 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, I𝑝𝑝, I𝑛𝑛} Train all the model if iteration > 200 & Similarity(I𝑝𝑝, I𝑡𝑡) > th: Insert tgt frame I𝑡𝑡 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset ℐ𝑠𝑠 {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} 𝐈𝐈𝑡𝑡 Model Loss Train
  • 32. 31 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, I𝑝𝑝, I𝑛𝑛} Train all the model if iteration > 200 & Similarity(I𝑝𝑝, I𝑡𝑡) > th: Insert tgt frame I𝑡𝑡 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset ℐ𝑠𝑠 {𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} 𝐈𝐈𝑡𝑡 Model Loss Train
  • 33. 32 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Train all the model if iteration > 200 & Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th: Insert tgt frame 𝐈𝐈𝒕𝒕 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset ℐ𝑠𝑠 𝐈𝐈𝑝𝑝 𝐈𝐈𝑡𝑡 Model Score
  • 34. 33 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Train all the model if iteration > 200 & Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th: Insert tgt frame 𝐈𝐈𝒕𝒕 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset ℐ𝑠𝑠 𝐈𝐈𝑝𝑝 𝐈𝐈𝑡𝑡 Model Score
  • 35. 34 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Train all the model if iteration > 200 & Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th: Insert tgt frame 𝐈𝐈𝒕𝒕 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset ℐ𝑠𝑠 𝐈𝐈𝑝𝑝 𝐈𝐈𝑡𝑡 Model Score
  • 36. 35 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Train all the model if iteration > 200 & Similarity(𝐈𝐈𝑝𝑝, 𝐈𝐈𝑡𝑡) > th: Insert tgt frame 𝐈𝐈𝒕𝒕 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset ℐ𝑠𝑠 𝐈𝐈𝑝𝑝 𝐈𝐈𝑡𝑡 Model Score
  • 37. 36 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Train all the model if iteration > 200 & Similarity(I𝑝𝑝, I𝑡𝑡) > th: Insert tgt frame I𝑡𝑡 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset 𝐈𝐈𝑛𝑛𝐈𝐈𝑝𝑝 Model ℐ𝑠𝑠 𝐈𝐈𝑡𝑡 Scores
  • 38. 37 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, 𝐈𝐈𝑝𝑝, 𝐈𝐈𝑛𝑛} Train all the model if iteration > 200 & Similarity(I𝑝𝑝, I𝑡𝑡) > th: Insert tgt frame I𝑡𝑡 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset Model
  • 39. 38 Proposed method • Training – Keyframe update & management Random KF initialization for epoch: for iteration: Choose training pair {ℐ𝑠𝑠, I𝑝𝑝, I𝑛𝑛} Train all the model if iteration > 200 & Similarity(𝐼𝐼𝑝𝑝, 𝐼𝐼𝑡𝑡) > th: Insert tgt frame 𝐼𝐼𝑡𝑡 as KF Merge KF KF pool 𝒫𝒫 𝐾𝐾 Dataset Model この操作を繰り返すことで KF poolの最適化を行う
  • 40. 39 Experimental results • KITTI dataset – Monocular Depth Estimation KF selectionによって学習データを調整することで, 学習が安定し 推定精度も高くなる
  • 41. 40 Experimental results • KITTI dataset – Monocular Depth Estimation KF selectionによって学習データを調整することで, 学習が安定し 推定精度も高くなる
  • 42. 41 Experimental results • KITTI dataset – Absolute Trajectory Error (ATE) KF selectionがdata augmentationの効果を持ち, 結果としてカメラ 姿勢の推定精度が向上
  • 43. 42 Experimental results • KITTI dataset – Average Rotation Errors とはいえカメラの回転の推定精度はORB-SLAM[Mur-Artal, TOR15]には 勝てていない状況
  • 44. 43 Experimental results • KITTI dataset – Keyframe selection • カメラが並進する場所では, 均一になるように選択 • カメラが回転する場所では, 変化が激しいのでより刻んだ選択
  • 45. 44 Experimental results • KITTI dataset – Ablation study Depth推定 カメラ軌跡推定
  • 46. 45 Conclusion • SfMLearner with KF selection – VSLAMで最も重要なKF selectionを, SfMLearnerの枠組みに追加 – UnsupervisedでKF selectionを学習する手法を提案 – 従来手法よりも高精度な奥行き推定, カメラ姿勢推定を達成. • 感想 – 従来人手の緻密な設計が必要だったKF selectionを, unsupervisedに CNNで学習し実現した点が新しく非常に面白い – KF selectionだけでなく, Bundle Adjustment等の最適化要素も追加 できるとDeep SLAMの実現により近付きそう