The aim of this work is to semi-automatise labelling image process. We want to achieve this through:
- speeding up the process
- decrease the human error
- make the process reproducible
Here we want to label a set of 1k images with some quality measures. Those labels are:
- overall quality
- light
- growth
- focus
im_id | overall | light | ... | focus |
---|---|---|---|---|
im1 | NA | NA | ... | NA |
im2 | NA | NA | ... | NA |
im3 | NA | NA | ... | NA |
- load a set of images (specify the path)
- visualise one image at time
- input the value and submit
- go to next image and repeat
- save and eventually continue from this step
im_id | overall | light | ... | focus |
---|---|---|---|---|
im1 | 0.1 | 0.1 | ... | 0.3 |
im2 | 0.9 | 0.6 | ... | 1 |
im3 | 0.7 | 0.6 | ... | 0.5 |
When the user defines the list of images a database have to be created. When the user submit his values, they should be added to the images on the db.
Some test images to play with can be found here
- Number and name of the metrics could vary depending on the need