Thank you! This is very useful in creating datasets to train. I've watched all of your videos on stylegan multiple times. Started from your google colab notebook, bought a rtx 3080 and now run it on my own PC. Your videos got me there. appreciate you!
Dear Sir, Appreciate your effort in putting this wonderful video. I believe you have helped us solve a major problem on how to get data sets. Keep us updated about the improvements to this code and any other utilities. I greatly appreciate your videos.
This video is very interesting. The only problem the author assumes that the user should know a lot about Python or Flickr because refers to many things like Anaconda, and other tools that is not common to the regular patron of this video.
Very nice, thanks. Certainly a lot simpler than gallery-dl! The only problem so far is I get very few images. For example, a search for “cats” in licenses 7,8,9,10 gives 17 images. A manual search shows many more cc0 and public domain images available?
Is it possible to be able to generate metadata for the set of images downloaded? I am interested in doing further analysis (location analysis) with the images data. I'm particularly interested in the geo information. Can such information be obtained through your code?
I think a nice feature would be max/min resolution file to keep. So, if a file that is downloaded is 100 x 100 and I have a min of 256 x 256, the file won't be kept.
Still relevant today! Thanks, the only thing to update would be the PIL.Image.ANTIALIAS attribute to Image.Resampling.LANCZOS in the flickr-download.py file
It would be great if you did a tutorial on how to use your trained model. Have been trying to use your pretrained 70's artwork colab book to load my own model, but it keeps getting stuck at the loading model step as you load from a URL where as I would be loading from google drive.
I would probably try pulling from gdrive to /content/, then load from there... pulling data across google drive in realtime has caused me timeout issues before.
I am having a small problem with the download script. When I search directly through the search box on flickr I get many images. But when I enter the same search query into the .ini file, I don't get any images (also when allowing all license types 0-10). Certain other queries work, certain not. I don't understand what could be the problem here? Any hints?
solved it myself by commenting out "tags=..." within the self.flickr.walk() call. Apparently, using "text=..." and "tags=..." being set to the same value was messing something up...
@@HeatonResearch Yeah it used to be cool but Flickr's technical side basically died in 2018. Maintaining it is basically just adjusting for Python versions. Kind of a death of Web 2.0 thing.
Thank you! This is very useful in creating datasets to train. I've watched all of your videos on stylegan multiple times. Started from your google colab notebook, bought a rtx 3080 and now run it on my own PC. Your videos got me there. appreciate you!
Thanks, this was extremely useful. Would love to see more of the utilities/pipeline!
Thanks, Jeff! Really appreciate this series on all these utilities you've saved me a ton of time.
Thanks a lot!
Actually getting into an evoked emotions project with CNN
Best from Argentina and don´t stop making this videos!!
I've been watching your videos all day. Thank you so much for sharing all your knowleadge and these amazing tools !
Glad you like them!
Just what I needed. Crawling Google is so useless when you can only get about 75 images. Thank you dear sir
You are most welcome
Please show more of your pipeline :D
yea that would be great - learning so much from jeff at the minute
Dear Jeff, you are a legend !
Thanks for the great content, greetings from Mauritius
definitely would like to see more pipeline details!
Dear Sir,
Appreciate your effort in putting this wonderful video. I believe you have helped us solve a major problem on how to get data sets.
Keep us updated about the improvements to this code and any other utilities.
I greatly appreciate your videos.
This video is very interesting. The only problem the author assumes that the user should know a lot about Python or Flickr because refers to many things like Anaconda, and other tools that is not common to the regular patron of this video.
Hey Jeff Awesome video really love the concept. You should make more computer vision videos.
Very nice, thanks. Certainly a lot simpler than gallery-dl! The only problem so far is I get very few images. For example, a search for “cats” in licenses 7,8,9,10 gives 17 images. A manual search shows many more cc0 and public domain images available?
This is fantastic thank you. I've like, subscribed and shared on linkedin
Great, thank you, I saw the Linked In post.
00:16 frog statues? no way in hell i would have ever predicted that was what you was gonna say ahha
This proved really helpful. Thanks!
Thank you! This is very useful! You have great videos.
Thankyou so much! saved a lot of time
Hi I would also like the video where you go through the pipeline - especially the logging.properties file
Is it possible to be able to generate metadata for the set of images downloaded? I am interested in doing further analysis (location analysis) with the images data. I'm particularly interested in the geo information. Can such information be obtained through your code?
I think a nice feature would be max/min resolution file to keep. So, if a file that is downloaded is 100 x 100 and I have a min of 256 x 256, the file won't be kept.
Hi, I dont get the flickrapi as a module in anaconda there is only r-flickrapi?
Frog statues :D
Still relevant today! Thanks, the only thing to update would be the PIL.Image.ANTIALIAS attribute to Image.Resampling.LANCZOS in the flickr-download.py file
Very nice
Please make some vedios on visual question answering
+1 for a video in which you show your process for downloading Google images.
It would be great if you did a tutorial on how to use your trained model. Have been trying to use your pretrained 70's artwork colab book to load my own model, but it keeps getting stuck at the loading model step as you load from a URL where as I would be loading from google drive.
I would probably try pulling from gdrive to /content/, then load from there... pulling data across google drive in realtime has caused me timeout issues before.
it shows a attribute error as xml doesnt have getchildren().please suggest some solutions
I am having a small problem with the download script. When I search directly through the search box on flickr I get many images. But when I enter the same search query into the .ini file, I don't get any images (also when allowing all license types 0-10). Certain other queries work, certain not. I don't understand what could be the problem here? Any hints?
solved it myself by commenting out "tags=..." within the self.flickr.walk() call. Apparently, using "text=..." and "tags=..." being set to the same value was messing something up...
Sir do a video on image super Resolution using SRGAN
You're awesome
I am getting error
It says keyerror : formatters
Hey , have you found a solution to this problem? Please do let me know. Thanks.
Let's hope flickr is around long enough. Maintaining the API as it ages is a pain.
Yes that is true, always breaking changes on these kinds of thing.
@@HeatonResearch Yeah it used to be cool but Flickr's technical side basically died in 2018. Maintaining it is basically just adjusting for Python versions. Kind of a death of Web 2.0 thing.
Show the entire pipeline
Yes, more will be coming soon.