X

Download Evaluating Sketch Query Interfaces PowerPoint Presentation


Login   OR  Register
X


Iframe embed code :



Presentation url :

X

Description :

Evaluating Sketch Query Interfaces powerpoint presentation, this ppt presentation evaluating sketch query interfaces for free.

Tags :

sketch query | 3d | 3d model | 3d tips | 3d tutorial

Home / Education & Training / Education & Training Presentations / Evaluating Sketch Query Interfaces PowerPoint Presentation

Evaluating Sketch Query Interfaces PowerPoint Presentation

Ppt Presentation Embed Code   Zoom Ppt Presentation

About This Presentation


Description : Evaluating Sketch Query Interfaces powerpoint presentation, this ppt presentation evaluating sketch ... Read More

Tags : sketch query | 3d | 3d model | 3d tips | 3d tutorial

Published on : Jun 26, 2014
Views : 340 | Downloads : 0


Download Now

Share on Social Media

             

User Presentation
SlidesFinder: Advertise with us
Related Presentation
No Other Uploads
Free PowerPoint Templates

Evaluating Sketch Query Interfaces
Presentation Transcript
Collapse/Expand

Slide 1 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001
Slide 2 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web
Slide 3 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query
Slide 4 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc.
Slide 5 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch
Slide 6 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results
Slide 7 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results
Slide 8 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries?
Slide 9 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method
Slide 10 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object
Slide 11 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection
Slide 12 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare
Slide 13 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix
Slide 14 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix
Slide 15 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix
Slide 16 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix
Slide 17 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3
Slide 18 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views:
Slide 19 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views:
Slide 20 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections
Slide 21 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections 1 vs. 2 vs. 3 Query Projections 3D
Slide 22 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections 1 vs. 2 vs. 3 Query Projections 3D 3D vs. Three Projections 3D Three 2D Projections
Slide 23 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections 1 vs. 2 vs. 3 Query Projections 3D 3D vs. Three Projections 3D Three 2D Projections 3D vs. Three Projections 3D Three 2D Projections
Slide 24 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections 1 vs. 2 vs. 3 Query Projections 3D 3D vs. Three Projections 3D Three 2D Projections 3D vs. Three Projections 3D Three 2D Projections Conclusion, Future work 2D outline query interface looks promising Different and/or additional input, 2D: skeletons silhouette edges Evaluation comparison with text query user study better model database
Slide 25 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections 1 vs. 2 vs. 3 Query Projections 3D 3D vs. Three Projections 3D Three 2D Projections 3D vs. Three Projections 3D Three 2D Projections Conclusion, Future work 2D outline query interface looks promising Different and/or additional input, 2D: skeletons silhouette edges Evaluation comparison with text query user study better model database Acknowledgements Princeton Shape Analysis Group csstaff@cs, CS department technical staff NSF http://shape.cs.princeton.edu/search
Slide 26 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections 1 vs. 2 vs. 3 Query Projections 3D 3D vs. Three Projections 3D Three 2D Projections 3D vs. Three Projections 3D Three 2D Projections Conclusion, Future work 2D outline query interface looks promising Different and/or additional input, 2D: skeletons silhouette edges Evaluation comparison with text query user study better model database Acknowledgements Princeton Shape Analysis Group csstaff@cs, CS department technical staff NSF http://shape.cs.princeton.edu/search Search Engine Website initial 48 hr. crawl resulted in 23,000+ VRML models 2,000 have been processed, about 20-25 % are single objects (some duplicates) rest is terrains, visualizations, VEs, too simple, molecules, etc. they are cached and available for download
Slide 27 - Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and Analysis of 3D Models Tuesday October 30, 2001 Goal Shape-based search engine for 3D models on Web 3D Model Search Engine Shape Signatures Analysis 3D Model Database Shape Signature Query Result Compare Analysis User Query Challenges Web crawling directed search Shape indexing shape distributions symmetry descriptors etc. Query interface text sketch etc. Sketch Query Interface Spectrum More descriptive Easier input 2D Sketch 3D Sketch 3D Sketch Example Results 2D Sketch Example Results Which Sketch Interface is “Better” User study (future) Log behavior of users with 2D and 3D interfaces Required to assess “ease of input” In preparation Initial study without users Use 3D models themselves as 3D queries Use projections of 3D models as 2D queries Compare discriminating power Can 2D queries be as discriminating as 3D queries? 133 models classified into 25 categories Same dataset as Osada et al. 4 Mugs 6 Cars 3 Boats Matching 3D Models with Shape Distributions Osada et al., I3D 2001 Evaluation Method 3D Comparison Method Database with D2 distributions Query D2 distribution Query Result Compare Query object 2D Comparison Method 2 1 3 4 7 6 5 7 Database projections per model Query projection 2D Comparison Method 2 1 3 4 7 6 5 7 Database D2 distributions per model Query 2D distribution Compare diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 3D Similarity Matrix diagonal nearest neighbor first tier first two tiers query models matches 2D Similarity Matrix Multiple 2D Projections Query 2 1 3 4 7 6 5 Database D2 distributions Query 2D distribution Compare 2 1 3 Example: 3 Projection Query stored views: query views: Example: 3 Projection Query minimize sum of similarity scores stored views: query views: 3D 1 vs. 2 Query Projections 1 vs. 2 vs. 3 Query Projections 3D 3D vs. Three Projections 3D Three 2D Projections 3D vs. Three Projections 3D Three 2D Projections Conclusion, Future work 2D outline query interface looks promising Different and/or additional input, 2D: skeletons silhouette edges Evaluation comparison with text query user study better model database Acknowledgements Princeton Shape Analysis Group csstaff@cs, CS department technical staff NSF http://shape.cs.princeton.edu/search Search Engine Website initial 48 hr. crawl resulted in 23,000+ VRML models 2,000 have been processed, about 20-25 % are single objects (some duplicates) rest is terrains, visualizations, VEs, too simple, molecules, etc. they are cached and available for download Thank You