Added nomask generator tool #28
Conversation
tejasmhos
requested review from
vivanvish,
havanagrawal,
pshivraj and
lmtoan
| @@ -0,0 +1,6 @@ | |||
| { | |||
There was a problem hiding this comment.
Can we not check in checkpoint files? .ipynb_checkpoints should be in .gitignore, please add it if it doesn't already exist.
| "metadata": {}, | ||
| "outputs": [], | ||
| "source": [ | ||
| "def create_image_dataset(search_term, number_of_images, augmentation=0):\n", |
There was a problem hiding this comment.
Can we use boolean variables (True/False) instead of 0/1 for flags?
There was a problem hiding this comment.
That's actually not a boolean, it's an integer specifying how many augmentations should be generated.
There was a problem hiding this comment.
Ah I see, my bad, two follow up points
- Can we rename the argument to indicate its a count of the number of augmented images generated?
- Why do we need to generate the augmented images? I believe it is good practice to generate these on the fly during training/testing (@pshivraj)
There was a problem hiding this comment.
I think its good to add augmentation at one place in order to have a better sense as to what's happening. I would also recommend using scripts which are importable rather notebooks for better readability and version control.
| " x = path.rstrip('\\n')\n", | ||
| " t = x.split('/')[-1].rstrip('.jpg')\n", | ||
| " im = Image.open(x)\n", | ||
| " im_size = im.size\n", |
There was a problem hiding this comment.
Is this really necessary? Why can't we just use im.size?
There was a problem hiding this comment.
Alright, I've made some changes, including modularizing the code so it can be imported into notebook, and also removing augmentation.
Added a tool to generate images from Google and create blank mask images for these images.