Processing
times
Information provided by the participants
|
Run |
Processing times |
Computers characteristics |
|
Run 1 |
low-level features extraction
: 6 sec / picture model learning : 832 sec semantic concepts
prediction : 0.1 sec / picture |
Pentium4, 2.8GHz, 2Go, Linux |
|
Run 2 |
low-level features extraction
: 6 sec / picture semantic concepts
prediction : 0.1 sec / picture |
Pentium4, 2.8GHz, 2Go, Linux |
|
Run 3 |
low-level features extraction
: 6 sec / picture model learning : 176 sec semantic concepts
prediction : 0.05 sec / picture |
Pentium4, 2.8GHz, 2Go, Linux |
|
Run 4 |
low-level
features extraction : 6 sec / picture model
learning : 104 sec semantic
concepts prediction : 0.03 sec /
picture |
Pentium4, 2.8GHz, 2Go, Linux |
|
Run 5 |
low-level features extraction
: 6 sec / picture model learning : 577 sec semantic concepts
prediction : 0.02 sec / picture |
Pentium4, 2.8GHz, 2Go, Linux |
|
Run 6 |
Features extraction :
0.33s/image Apprentissage : global 900
secondes Retrieval : 295 seconde
/ requête |
Pentium4, 3.2GHz, 512Mo, Linux |
|
Run 7 |
about 15min for each class (1) pre-processing: 1 or 2s / image. (2) Features (color) extraction : 1s / image (3) Features (texture) extraction : 3/4s / image (4) Merging
attributes : ~ 1h i.e 5s / image. Learning : ~ 2 / 3h. Retrieval : ~ 1 min /
query |
Xeon 3Ghz, 2Go |
|
Run 8 |
3h for supervised clustering,
and about 5min for each class (1) pre-processing: 1 or 2s / image. (2) Features (color) extraction : 1s / image (3) Features (texture) extraction : 3/4s / image (4) Merging
attributes : ~ 1h i.e 5s / image. Learning : ~3h Retrieval : ~5 min /
query |
Xeon 3Ghz, 2Go |
|
Run 9 |
Features extraction :0.61
s/image Global learning : 11160 s Retrieval : 118 s/query |
Pentium 4, 2.80 GHz, 512 Mo,
Linux |
|
Run 10 |
Indexing / Features extraction
: 3 sec / image Learning : 20 min / training
set (5000 images) Research / Classification
(creation of the results) : 0.22 sec
/ image |
Laptop, Pentium M, 1.60Ghz,
512Mo, Matlab, Win, Linux |
|
Run 11 |
Indexing / Features extraction
: 3 sec / image Learning : 20 min / training
set (5000 images) Research / Classification
(creation of the results) : 0.22 sec
/ image |
Laptop, Pentium M, 1.60Ghz,
512Mo, Matlab, Win, Linux |
|
Run 12 |
Indexing / Features extraction
: Visual extraction and indexing :
1h40 for train + test images (no dev) Learning : 2 hours 10 minutes,
training of an ANN with TORCH toolbox Research / Classification
(creation of the results) : around 3 minutes for the ANN
forward (~50 inputs, 70 nhu, 11 ouputs) Global processing : ~6h
without parallel processing Extra time : fusion
(intersection) for estimates of other 2 classes and formating results to Trec : 30 minutes for the final work. |
Bi-Xéon 3GHz, 4Go, Linux +
Matlab |
MAP
+ Retrieval time (second / concept) log10 scale
MAP
+ Features extraction time (second/image)
MAP
+ Model learning (global time for all the concepts)
