Pollenometer
Jump to navigation
Jump to search
Jim and Greg : April 2008
Current status:
- Classifypollen
- The Bad News
- Clutter is exclude on the basis of size, but this means very big clutter is declared "pine"
- The Good News
- Tiny pine in 01-08-08 was called "clutter" by the program
- We think the program is right!
- The Bad News
- Where do we go from here
- Make a pretty movie!
- Run counts on 01-08-08 and compare to Jim
- Ditto for 01-01-08
| Database | Manual Count | Manual Image-based Count | Automatic Count |
| 01-01-08 | |||
| 01-08-08 | |||
| 01-15-08 | |||
| etc. |
Jim and Greg : March 2009
It's time for large-scale benchmarking!
02-19-09
| Class | Linecount | LabelGUI | Computer v1 |
Computer v2 |
Computer v3 |
| alder | 2 | 1 | |||
| ash+privet | 4+6 | 3 | |||
| birch | 3 | 2 | |||
| cypress | 5 | 4 | 5 | 4 | 3 |
| eucalyp | 3 | 1 | |||
| grass | 2 | NA | 3 | 1 | |
| olive | 1 | 1 | 1 | ||
| oak | 3 | 1 | 1 | 1 | |
| palm | 4 | 1 | |||
| pine | 13 | 7 | 16 | 17 | 6 |
| poplar+cottenwood | 5 | 4 | |||
| sycam=planetree | 3 | 6 | 6 | 6 | 3 |
| unknown | 2 | 8 | NA |
v2: ntrain_bias=0.5; shape_bias=0.2; clutter_bias=0.0;
v3: ntrain_bias=0.5; shape_bias=0.2; clutter_bias=-0.0
Here is a more complete record of the
options that are in use
for the above classifypollen tests.
02-22-09
4-23-09
| Class | Linecount | LabelGUI | Computer v1 |
Computer v2 |
Computer v3 |
| alder | |||||
| ash+privet | |||||
| birch | 2 | ||||
| chenopod | 1 | ||||
| cypress | 1 | ||||
| eucalyp | 1 | 41 | |||
| ginko | 8 | ||||
| grass | 1 | 1 | |||
| olive | |||||
| oak | 4 | 9 | |||
| palm | |||||
| pine | 1 | 9 | |||
| poplar+cottenwood | |||||
| sage | 1 | 1 | |||
| sycam=planetree | 1 | ||||
| unknown |
v1
Here is a more complete record of the
options that are in use
for the above classifypollen tests.
4-26-09
| Class | Linecount | LabelGUI | Computer v1 |
Computer v2 |
Computer v3 |
| alder | |||||
| ash+privet | |||||
| birch | |||||
| cypress | |||||
| eucalyp | |||||
| grass | |||||
| olive | |||||
| oak | |||||
| palm | |||||
| pine | |||||
| poplar+cottenwood | |||||
| sycam=planetree | |||||
| unknown |
v1
Here is a more complete record of the
options that are in use
for the above classifypollen tests.
5-29-09
| Class | Linecount | LabelGUI | Computer |
| crepe | 1 | ||
| eucalyp | 1 | ||
| gingko | 4 | 16 | |
| grass | 5 | ||
| mulberry | 1 | ||
| pine | 7 | 2 | 4 |
| poplar+cottenwood | 5 | 1 | |
| sageb | 17 | 18 | 12 |
| sycam=planetree | 6 | 2 | |
| unknown | 5 | 12 |
Old Links
Ruxandra's log | Pollenometer components | Data=
References
- M. Ranzato , P.E. Taylor, J.M. House, R.C. Flagan, Y. LeCun P. Perona. Automatic Recognition of Biological Particles in Microscopic Images. Pattern Recognition Letters. PDF