happycoding

RT.

A wired problem I found when testing the VOC dataset 2007

1

``` def get_conv_image_descriptor_for_image(image, model): im = cv2.resize(image, (224, 224)).astype(np.float32) print('resized: {}'.format(im.shape)) dim_ordering = K.image_dim_ordering() if dim_ordering == 'th': # 'RGB'->'BGR' im = im[::-1, :, :] # Zero-center by mean pixel...

The results on Pascal-voc cannot be reproduced

3

Hi, Thanks for your excellent work. I have tried reproducing the results on PASCAL-VOC, but inferior results are obtained: **The model trained with CIRKD (72.6) is even 1% mIoU LOWER...

As I noticed, many other published models were built on VGG16. Have you tried applying VGG16 to your framework ? I think resnet50 can help get better performance when compared...

Could you show some result figures to prove its effectiveness ? Thanks ~

Fail to reproduce the results of GZLSS / GFLSS on PASCAL VOC

13

I have trouble in setting up the experimental environment on the public server, so I use a stronger baseline that can reach mIoU 59.8 after fine-tuning on novel classes with...