search

3neutronstar / GuidedMixup

Official PyTorch implementation of "GuidedMixup An Efficient Mixup Strategy Guided by Saliency Maps" (AAAI'23 Oral)
GNU General Public License v3.0
5 stars 4 forks source link

GuidedMixup Output #3

Closed khawar-islam closed 10 months ago

khawar-islam commented 10 months ago

Dear @3neutronstar

Thanks for your work. I am quite interested to visualize GuidedMixup images, to achieve this, I have added save_image(input_var[0], f'image_epoch_{epoch}_batch_{i}.png', args.std, args.mean) before prec1, prec5 = accuracy(output, target, topk=(1, 5)) in main.py file but the output is similar to original image. image_epoch_0_batch_0 image_epoch_0_batch_1

3neutronstar commented 10 months ago

Dear @khawar-islam

Thanks for your interest in my work. To visualize the GuidedMixup images, I recommend referring to the following permalink. Uncomment the line, and you can get the visualized images. link

Because the variable named input_var is not changed by using any mixup algorithm described in main.py, your results are the natural outcome.

Again, thank you for your interest in our work.