The mask used for testing

[cmyyy]

Hello, my question is what does the mask used for testing on places2 , celebaHQ-256 looks like when you do quantitative experiment?

[shepnerd]

In our paper, we used rectangle masks in random locations for testing. The released code and some pretrained models support the evaluation on the stroke masks.

cmyyy [email protected] 于2019年7月18日周四 上午10:55写道：

Hello, my question is what does the mask used for testing on places2 , celebaHQ-256 looks like?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/shepnerd/inpainting_gmcnn/issues/25?email_source=notifications&email_token=ABGU6CH2SEC6TN4M3SCD5JDP77LRHA5CNFSM4IEWPEU2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4G74W3NA, or mute the thread https://github.com/notifications/unsubscribe-auth/ABGU6CH6Q6QH7ZLP3JPHBUTP77LRHANCNFSM4IEWPEUQ .

[cmyyy]

So one 128*128 rectangle mask at random location of an image, and it is same for both the datasets,right?

[shepnerd]

Yes.

cmyyy [email protected] 于2019年7月18日周四 下午1:19写道：

So one 128*128 rectangle mask at random location of an image, and it is same for both the datasets,right?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/shepnerd/inpainting_gmcnn/issues/25?email_source=notifications&email_token=ABGU6CEH26Y5W52VIC7EZJTP774OTA5CNFSM4IEWPEU2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2HLI5Q#issuecomment-512668790, or mute the thread https://github.com/notifications/unsubscribe-auth/ABGU6CBRF5MWUPIRBXHM6B3P774OTANCNFSM4IEWPEUQ .

[cmyyy]

And when do quantitative experiment on places2, do u randomly chose 2000 images? If so, shape will be different, and in the test.py "input_image_tf = tf.placeholder(dtype=tf.float32, shape=[1, config.img_shapes[0], config.img_shapes[1], 3])" the shape is fixed, how do you cope with this situation?

[shepnerd]

Yes. 2k images are randomly chosen for evaluation since the full testing set of places2 is huge. About the shape problem, L63-70 in test.py just crop and resize the central part of the input images into the ones with the target image size, which ensures the input images to the network are with the same size.

cmyyy [email protected] 于2019年7月18日周四 下午1:58写道：

And when do quantitative experiment on places2, do u randomly chose 2000 images? If so, shape will be different, and in the test.py "input_image_tf = tf.placeholder(dtype=tf.float32, shape=[1, config.img_shapes[0], config.img_shapes[1], 3])" the shape is fixed, how do you cope with this situation?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/shepnerd/inpainting_gmcnn/issues/25?email_source=notifications&email_token=ABGU6CC7ELWZ24YSOVJWXLDQAABA5A5CNFSM4IEWPEU2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2HNC2A#issuecomment-512676200, or mute the thread https://github.com/notifications/unsubscribe-auth/ABGU6CG5SNEPY2J7P5RWEFTQAABA5ANCNFSM4IEWPEUQ .

[cmyyy]

Another question,in the quanttative experiment, the places2 model was trained on random strokes or rectangle masks?

[shepnerd]

All experiments in the paper are conducted with rectangle masks.

cmyyy [email protected] 于2019年7月20日周六 下午3:46写道：

Another question,in the quanttative experiment, the places2 model was trained on random strokes or rectangle masks?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/shepnerd/inpainting_gmcnn/issues/25?email_source=notifications&email_token=ABGU6CEQGQYRRYJVZH6FVO3QAK7FXA5CNFSM4IEWPEU2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2NJAUQ#issuecomment-513445970, or mute the thread https://github.com/notifications/unsubscribe-auth/ABGU6CANLRCERXNGQ4DWGSLQAK7FXANCNFSM4IEWPEUQ .

[cmyyy]

Could you provide the places2 pretrained model trained on rectangle masks?

[shepnerd]

https://github.com/shepnerd/inpainting_gmcnn/issues/20#issuecomment-506572261

[godchengzhihang]

And when do quantitative experiment on places2, do u randomly chose 2000 images? If so, shape will be different, and in the test.py "input_image_tf = tf.placeholder(dtype=tf.float32, shape=[1, config.img_shapes[0], config.img_shapes[1], 3])" the shape is fixed, how do you cope with this situation?

Hello, is there a tensorflow version of the code for the quantitative analysis part of the paper?