Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
[RSJ22] TDP-MAT: Multimodal Language Comprehens...
Search
Semantic Machine Intelligence Lab., Keio Univ.
PRO
September 05, 2022
Technology
850
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
[RSJ22] TDP-MAT: Multimodal Language Comprehension for Object Manipulation Tasks via Real Images
Semantic Machine Intelligence Lab., Keio Univ.
PRO
September 05, 2022
More Decks by Semantic Machine Intelligence Lab., Keio Univ.
See All by Semantic Machine Intelligence Lab., Keio Univ.
[RSJ26] Building a VLA Model Based on Self-Distilled Classification
keio_smilab
PRO
0
190
[RSJ26] AnoleVLA: Lightweight Vision-Language-Action Model with Deep State Space Models for Mobile Manipulation
keio_smilab
PRO
0
120
[RSJ26] NarrativeFlow: Flow-Based Vision-Language-Action Model Using Robot Velocity Fields
keio_smilab
PRO
0
150
[RSJ26] Flow as Flow: Modeling Robot Velocity Fields as Probability Velocity Fields
keio_smilab
PRO
0
190
[RSJ26] Hierarchy-Aware Multimodal Retrieval-Augmented Generation for Embodied Question Answering
keio_smilab
PRO
1
170
[MIRU26] Open-Vocabulary Intention-Guided Object Detection in Diverse Scenes
keio_smilab
PRO
0
190
[Journal club] Evo-1: Lightweight Vision-Language-Action Model with Preserved Semantic Alignment
keio_smilab
PRO
0
48
[MIRU26] To What Extent Does MLLM-as-a-Judge Exhibit Cross-Model Preference Bias?
keio_smilab
PRO
0
280
[Journal club] FlashVID: Efficient Video Large Language Models via Training-free Tree-based Spatiotemporal Token Merging
keio_smilab
PRO
0
36
Other Decks in Technology
See All in Technology
Azure App Service / Container Apps の組み込み認証
kuniteru
0
190
深夜のクラウド懺悔室 1:29:300 or 1:0:0
kazzpapa3
0
150
ブラウザアプリの継続的パフォーマンスモニタリング (序) / Continuous Browser Application Performance Monitoring; Act 1
moznion
0
110
データエンジニアリングワークショップ:Auto LoaderとSparkで学ぶデータパイプライン構築
databricksjapan
PRO
0
150
Level Up Your CDK DX: 5 Tools I’ve Been Building
gotok365
2
190
Jetpack Compose で挑む新聞紙面UI ─ 複合ジェスチャー・ポリゴン記事領域・適応的ページ構成という3つの壁/droidkaigi2026
nikkei_engineer_recruiting
0
190
【試作】IoT x AIエージェント
happysamurai294
0
100
AIで実装は速くなった。なのにプロダクトは速くならない。職能の壁を越えて価値のフローを設計する
nwiizo
6
6.9k
[DroidKaigi 2026] Making UI specifications visible: Android UI development in the AI agent era supported by Compose Screenshot Testing and galleries
syarihu
0
550
設計の世代交代を乗り越える、14年続くAndroidアプリの開発戦略
sansantech
PRO
1
120
Webとヘルスデータ
yukukotani
0
170
AIとペアプロを始める。人とのペアプロをやめる。ペアプロの良さを改めて知る。もっと好きになった。 / Rediscovering Pair Programming
honyanya
1
200
Featured
See All Featured
The Organizational Zoo: Understanding Human Behavior Agility Through Metaphoric Constructive Conversations (based on the works of Arthur Shelley, Ph.D)
kimpetersen
PRO
0
430
YesSQL, Process and Tooling at Scale
rocio
174
15k
Evolution of real-time – Irina Nazarova, EuRuKo, 2024
irinanazarova
9
1.5k
How GitHub (no longer) Works
holman
316
150k
The Director’s Chair: Orchestrating AI for Truly Effective Learning
tmiket
1
290
The Cult of Friendly URLs
andyhume
79
7k
First, design no harm
axbom
PRO
2
1.3k
Exploring anti-patterns in Rails
aemeredith
3
490
WENDY [Excerpt]
tessaabrams
12
39k
How Fast Is Fast Enough? [PerfNow 2025]
tammyeverts
3
870
How People are Using Generative and Agentic AI to Supercharge Their Products, Projects, Services and Value Streams Today
helenjbeal
1
300
Building Adaptive Systems
keathley
44
3.2k
Transcript
1
2
3 ✓ https://www.toyota.com/usa/toyota-effect/romy-robot
✓ : 4 “Look in the left wicker vase that
is next to the potted plant” Wicker vase :
✓ : “Look in the left wicker vase that is
next to the potted plant” 5 Wicker vase : Wicker vase Wicker vase Wicker vase
✓ : ✓ Key : “Look in the left wicker
vase that is next to the potted plant” 6 Wicker vase : Wicker vase Wicker vase Wicker vase
✓ REVERIE-fetch • 7 “Look in the left wicker vase
that is next to the potted plant”
✓ REVERIE-fetch • • (Instruction) (Context Regions) (Candidate Region) 8
“Look in the left wicker vase that is next to the potted plant”
✓ REVERIE-fetch • • (Instruction) (Context Regions) (Candidate Region) 9
“Look in the left wicker vase that is next to the potted plant”
✓ REVERIE-fetch • • (Instruction) (Context Regions) (Candidate Region) •
10 “Look in the left wicker vase that is next to the potted plant”
✓ REVERIE-fetch • • (Instruction) (Context Regions) (Candidate Region) •
11 “Look in the left wicker vase that is next to the potted plant” Faster R-CNN[Ren+, PAMI16]
MTCM [Magassouba+, RA-L19] . VGG16LSTM . Target-dependent UNITER (TDU) [Ishikawa+,
RA-L21] UNITER[Chen+, ECCV20] . REVERIE task / dataset [Qi+, CVPR20] , REVERIE 12
• MAT[Ishikawa+, ICPR22] • CLIP[Radford+, ICML21] • Perceiver[Jaegle+, ICML21] 13
• MAT[Ishikawa+, ICPR22] • CLIP[Radford+, ICML21] • Perceiver[Jaegle+, ICML21] 14
• MAT[Ishikawa+, ICPR22] • CLIP[Radford+, ICML21] • Perceiver[Jaegle+, ICML21] 15
• MAT[Ishikawa+, ICPR22] • CLIP[Radford+, ICML21] • Perceiver[Jaegle+, ICML21] 16
• MAT[Ishikawa+, ICPR22] • CLIP[Radford+, ICML21] • Perceiver[Jaegle+, ICML21] 17
2 1 3
✓ 𝜹𝑡 ✓ 18 Input 𝜹𝑡 Output 1. 𝐸 𝜹
= CE 𝑓 𝒙 , 𝒚 ∇𝜹 𝐸 𝜹 = 𝜕𝐸 𝜕𝜹 2. ∇𝜹 𝐸 𝜹 𝒎𝑡 𝒗𝑡 𝒎𝑡 = 𝜌1 𝒎𝑡−1 + 1 − 𝜌1 ∇𝜹 𝐸 𝜹𝑡 𝒗𝑡 = 𝜌2 𝒗𝑡−1 + 1 − 𝜌2 ∇𝜹 𝐸 𝜹𝑡 2 3. 𝒎𝑡 𝒗𝑡 ∆𝜹𝒕 ෝ 𝒎𝑡 = 𝒎𝑡 1 − 𝜌1 𝑡 , ෝ 𝒗𝑡 = 𝒗𝑡 1 − 𝜌2 𝑡 ∆𝜹𝒕 = 𝜂 ෝ 𝒎𝑡 ෝ 𝒗𝑡 + 𝜖 4. 𝜹𝑡+1 = Π 𝜹 ≤𝜖 𝜹𝑡 + ∆𝜹𝒕 ∆𝜹𝒕 𝐹
✓ CLIP ✓ ViT[Dosovitskiy+, ICLR21] ✓ transformer [EOT] 19 [EOT]
✓ ✓ Perceiver CLIP 20 CLIP Encoders
✓ CLIP Encoders , Perceiver 21
✓ REVERIE-fetch dataset - REVERIE dataset ✓ REVERIE[Qi+, CVPR18] -
→ 1. , 2. https://yuankaiqi.github.io/REVERIE_Challenge/static/img/demo.gif 22 Matterport3D
✓ REVERIE-fetch dataset - REVERIE dataset ✓ REVERIE[Qi+, CVPR18] :
+ 23 , ↓ - REVERIE - - https://yuankaiqi.github.io/REVERIE_Challenge/static/img/demo.gif
✓ REVERIE-fetch dataset • REVERIE dataset #Samples Vocabulary size Average
sentence length 30532 2853 19.1 Training Validation Test 26808 2552 1172 24 “Look in the left wicker vase that is next to the potted plant”
“Go into the living room and give me the pillow
on the couch nearest the plant” 25 • → TDP-MAT
26 • → TDP-MAT ✓ Bounding box “Make haste to
the office and fluff the pillow sitting on the left of the chair”
• Acc [%] : 27 Condition Acc [%] ↑ Baseline
: TDU [Ishikawa+, IROS21] 73.3 0.485 Ours : TDP-MAT W/o MAT 72.5 3.55 W/o MAT + Smaller learning rate 74.4 0.831 W/o CLIP & Perceiver 74.1 1.47 W/o Pretraining 73.1 2.24 Full 75.3 0.691 +2.0
28 Condition Acc [%] ↑ Baseline : TDU [Ishikawa+, IROS21]
73.3 0.485 Ours : TDP-MAT W/o MAT 72.5 3.55 W/o MAT + Smaller learning rate 74.4 0.831 W/o CLIP & Perceiver 74.1 1.47 W/o Pretraining 73.1 2.24 Full 75.3 0.691 +2.8 - - 5 - ( ) - Smaller learning rate : 1/8 -
29 Condition Acc [%] ↑ Baseline : TDU [Ishikawa+, IROS21]
73.3 0.485 Ours : TDP-MAT W/o MAT 72.5 3.55 W/o MAT + Smaller learning rate 74.4 0.831 W/o CLIP & Perceiver 74.1 1.47 W/o Pretraining 73.1 2.24 Full 75.3 0.691 +1.2 - CLIP Encoders, Perceiver Module, - Cross Attention
30 Condition Acc [%] ↑ Baseline : TDU [Ishikawa+, IROS21]
73.3 0.485 Ours : TDP-MAT W/o MAT 72.5 3.55 W/o MAT + Smaller learning rate 74.4 0.831 W/o CLIP & Perceiver 74.1 1.47 W/o Pretraining 73.1 2.24 Full 75.3 0.691 +2.2 - TDU
✓ • ✓ • MAT • ✓ • 31
✓ ✓ 𝐿 𝑁 𝑅𝐿×𝐷 𝑅𝑁×𝐸 𝑅𝐿×𝐷, 𝑅𝑁×𝐷 → 𝑅𝐿×𝑁
𝑅𝐿×𝐷 𝑅𝐿×𝐷, 𝑅𝐿×𝐷 → 𝑅𝐿×𝐿 32
✓ ✓ ✓ ✓ 33
✓ 34 8 × 10−4 𝛽1 = 0.9, 𝛽2 =
0.99
✓ ✓ ✓ 35 19+6=25