Skip to content

Commit

Permalink
update
Browse files Browse the repository at this point in the history
  • Loading branch information
theyoucheng committed Oct 25, 2021
1 parent a1c18c7 commit 02c060b
Showing 1 changed file with 18 additions and 18 deletions.
36 changes: 18 additions & 18 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -39,41 +39,41 @@ usage: deepcover.py [-h] [--model MODEL] [--inputs DIR] [--outputs DIR]
```
python ./sfl-src/sfl.py --mobilenet-model --inputs data/panda --outputs outs --testgen-size 200
```
`--mobilenet-model` pre-trained keras model
`--inputs` input images folder
`--outputs` output images folder
`--testgen-size` the number of input mutants to generate for explaining (by default, it is 2,000)
`--mobilenet-model` pre-trained keras model

`--inputs` input images folder

`--outputs` output images folder

`--testgen-size` the number of input mutants to generate for explaining (by default, it is 2,000)


## More options
```
python src/deepcover.py --mobilenet-model --inputs data/panda/ --outputs outs --measures tarantula zoltar --x-verbosity 1 --masking-value 0
```
`--measures` to specify the SFL measures for explaining: tarantula, zoltar, ochiai, wong-ii
`--x-verbosity` to control the verbosity level of the explanation results
`--masking-value` to control the masking color for mutating the input image
`--measures` to specify the SFL measures for explaining: tarantula, zoltar, ochiai, wong-ii

`--x-verbosity` to control the verbosity level of the explanation results

`--masking-value` to control the masking color for mutating the input image


## To start running the causal theory based explaining:
```
python ./sfl-src/sfl.py --mobilenet-model --inputs data/panda --outputs outs --causal --testgen-iterations 50
```
`--causal` to trigger the causal explanation
`--testgen-iterations` number of individual causal refinement calls; by default, it’s 1
`--causal` to trigger the causal explanation

`--testgen-iterations` number of individual causal refinement calls; by default, it’s 1

## To load your own model
```
python src/deepcover.py --model models/gtsrb_backdoor.h5 --input-rows 32 --input-cols 32 --inputs data/gtsrb/ --outputs outs
```
`--input-rows` row numebr for the input image
`--input-cols` column numebr for the input image
`--input-rows` row numebr for the input image

`--input-cols` column numebr for the input image



Expand Down

0 comments on commit 02c060b

Please sign in to comment.