Picblock fix

[Mopar]

Sorry for not specifying, those were all problems with Nudity Blocker.

I think it is another variant of cached images because it refuses to analyze the images again. With the mozilla URL, none of the images are analyzed at all.

The issue with Google applies to both extensions. Yes, images are pixelated in the search results but when you click a particular image, it opens the black info pane. Images appear pixelated at first, then are replaced with the real image.

Then with Nudity Blocker, it has a problem with Google's auto loading the next page of images so every odd set of images is still blurred and every even set is analyzed.

https://github.com/rob204/Web-Cleaner-C ... ter/Source

Some of the images from the GitHub link do not load in Nudity Blocker.

[1azzu1]

this sounds very interesting... i hope you guys make it work :)

[1azzu1]

this might be a dumb question (i know nothing about coding)
but wouldn't it be easier to make something the scans the screen rather then the images themself?

[rob204]

I think it is another variant of cached images because it refuses to analyze the images again.

The issue with Google applies to both extensions. Yes, images are pixelated in the search results but when you click a particular image, it opens the black info pane. Images appear pixelated at first, then are replaced with the real image.

I suppose these statements belong together? Good that you remind me google has this black info pane thing. I haven't seen it in years because it apparently depends on which google domain you use.

Well, of course it doesn't work with Nudity Blocker. When you open an image in the info pane at first it shows the thumbnail version. Then when the original version finished downloading in the background it replaces the thumbnail version. Nudity Blocker doesn't check for these details. Once an image is checked it's checked. Should be fixed now, but it seems to be crashing at about 170 background threads. Maybe we need to create some kind of worker pool? Nobody I know could run 170 threads in parallel anyway.

On the other hand Web Cleaner detects this change without problem. But unfortunately it fails to download the original version. Nudity Blocker can load about any image from any website because it manipulates the http headers. But Nudity Blocker even fails when google runs on https while the image gets served from an http website. I think that is because we run the XMLHttpRequest in the content script. Maybe it would help if we move those tricky downloads to the central background page process. Or even manipulate the http headers like Nudity Blocker does. At least there should be a setting at the options page.

Then with Nudity Blocker, it has a problem with Google's auto loading the next page of images so every odd set of images is still blurred and every even set is analyzed.

Yeah, I've seen that too and didn't know why. I really thought you could rely on the image load even at least to some extent. But it seems like you can't. I've tested it now. Even though image src is set the whole time the onload event doesn't fire reliably when it transitions from not complete to complete. There's no other way to check for image loading other than checking every now and then.

With the mozilla URL, none of the images are analyzed at all.

https://github.com/rob204/Web-Cleaner-C ... ter/Source

Some of the images from the GitHub link do not load in Nudity Blocker.

Same problem with twitter. They use content security policies in the http header to restrict what kind of things you can do on their website. For example they prevent the execution Nudity Blocker's background worker. Web Cleaner doesn't have a problem with this because it runs the image analysis on the main thread. But again we could manipulate the headers or put it into the background page.

but wouldn't it be easier to make something the scans the screen rather then the images themself?

Hmm.. Yes, that's a good question. Actually I already hinted at this alternative. But it's a very different strategy. Something like that could probably not run as a browser extension. You would either have to compile it into the browser or as a separate application. And then I asked why to stop there. Instead of just concentrating on the browser you could have your whole operating system interface behind such a filter.

But currently we only concentrate on single still images. They get either blocked or not blocked. And I think it would be pointless to block the whole screen. So instead we would have to start thinking about how to separate the screen into regions of nudity and about scales and about how many pixels around those medium-tanned-skin-colored ones we want to block. There is no scaling information when you scan the whole screen. When you scan single images you can say more than 15% nudity is not permitted. But when you scan the whole screen then you don't know 15% of what? Should a big image get blocked while a small one does not?

And then we have to decide how to work with changes. With still images it's relatively easy because they don't change that much. But when you scroll a web page stuff appears and disappears. Sometimes it's only visible in parts. We would have to detect those changes, analyze and re-separate the screen into regions. Preferably in real-time.

Well, I think that would be a cool undertaking. I think the biggest problem would be to extract the scaling information. Certainly there must be someone who invented some clever algorithms for this problem already. And it would improve single image analysis as well to separate things and have scaling information.

[Mopar]

I suppose these statements belong together? Good that you remind me google has this black info pane thing. I haven't seen it in years because it apparently depends on which google domain you use.

No, they were not meant to be together.

Check out the image cache issue with this link.

http://imgbox.com/g/G3Aau5RqQs

Whichever process will make things easier or get everything to work. I don't have any knowledge in that.

[rob204]

Interesting, for me it works with both extensions. The thumbnails are visible but the big images are pixelated. It even works when both extensions are running at the same time even though that's a bad idea. There's only a problem with the ads sometimes.

Maybe you use an old version of the extension? I still haven't build in a way to auto-update them.

[YRD]

Wow nice work already guys. works like a charm.


Although, Chrome blocks the extension, because it's not from the store. So there is a fix.

1. If you delete the extension in the extension tab
2. Select developer mode
3. press 'load unpacked expansion'
4. select the source filepath of the webcleaner downloaded here

https://github.com/rob204/Web-Cleaner-C ... ter/Source