summaryrefslogtreecommitdiff
path: root/neural_style.py
AgeCommit message (Collapse)Author
2016-12-04Python 3 compatibilityJonathan Böcker
Map objects is not subscriptable in Python 3
2016-11-16Update neural_style.pyCameron
2016-11-16Update neural_style.pyCameron
2016-11-09Update neural_style.pyCameron
2016-11-07Update neural_style.pyCameron
2016-11-06Fixed commmit.cam
2016-11-06Protect from div by 0cam
2016-11-06Update neural_style.pyCameron
2016-11-06Update neural_style.pyCameron
2016-10-24Added error checking for input images as per request.cam
2016-10-22Added LAB colorspacecam
2016-10-22Simplified color conversioncam
2016-10-22Added more colorspaces for color conversioncam
2016-10-20Changed default style weight to 1e4Cameron
2016-10-17Added multiple content layerscam
2016-10-15Fixed bug between tensorflow builds 0.9, 0.10rc, 0.10cam
2016-10-15Fixed typoscam
2016-10-13Update neural_style.pydeepxiangfa
fix the tv_loss bug
2016-10-12Change interpolation method from bicubic to area.gtoderici
Change interpolation method from bicubic to area. This is necessary because when downsizing high resolution images, the bicubic algorithm generates high frequency noise (for example, try resizing an image containing gravel with bicubic vs. area). I guess ideally there should be a check to determine whether it's upsampling/downsampling and choose Lanczos for upsampling. However, given how slow this is... maybe we only need to worry about downsampling?
2016-10-11Added verbose to usage scriptCameron
2016-10-11Fixed tab indentationCameron
2016-10-11Add files via uploadCameron
2016-10-11Modified READMEcam
2016-10-10Modified READMEcam
2016-10-08Fixed 'style_mask' errorCameron
2016-10-08Fixed 'verbose' errorCameron
2016-10-08Fixed 'verbose' errorCameron
2016-10-08Add files via uploadCameron
2016-10-08First commitcam
generated by cgit v1.2.3-70-g09d2 (git 2.51.0) at 2025-12-10 13:56:32 +0000