sirfz

I don't personally visit twitch.tv but I just did and it loaded normally and streaming auto-played without enabling Flash (although the site requested it for some reason).

Speaking of Flash, some sites request Flash on Firefox but not on Chrome for example (although it's totally unnecessary). This is becoming less frequent recently but it shows that a lot of websites are not taking advantage of Firefox's full features as they do with Chrome (something to consider as well).

What does "faster" mean? honest question. I've been using Firefox since its early days and never jumped on the Chrome bandwagon simply because I didn't see the need to do so. When Firefox felt slow, it usually was the whole operating system that's slow and neither Chrome or any other browser could solve that. Nowadays, all the machines I use are powerful enough to not feel any such slowness and I just don't understand what is this "speed" people keep attributing to Chrome and what is so "slow" in Firefox (regardless of that fact they're working on speeding it up).

In addition, Chrome recently crashed on me multiple times while using Google Spreadsheet which is supposed to work better with Chrome than other browsers.

Same here, if I start needing to scroll through open tabs it means I should immediately start closing some.

Note that pytesseract has no bindings with tesseract's API, it merely uses the tesseract command line tool and communicates with it via stdout/stdin.

If you want native and complete access to tesseract's API you can use tesserocr: https://github.com/sirfz/tesserocr

If you're into huge laptops, I'm running Linux (Ubuntu) on an ASUS ROG G752VS right now, previously ran on G750 and G751 models. Build quality is solid and the specs are no doubt high end. However, not everything works out-of-the-box and some things have no fix as far as I know. For example with my current G752VS, the microphone port is completely useless as well as the function keys (only volume control Fn keys work). Other fixable issues are brightness control (fixed using xbacklight) and headphones jack (fixed using hdajackretask).

All changes are very well documented for each release [1]. The biggest changes I've seen were, as expected, upgrading from 0.x to 1.x which were mostly function attribute names changes that were made to comply with numpy's attribute names (a welcome change). They did provide tools that automatically convert 0.x code to 1.x so that was helpful.

It is not necessary to upgrade if you're satisfied with the current version and I sure won't deploy a different version to production than training. Upgrading is mainly worth it if there are performance improvements or new features/tools that make a big impact (and those are the changes we're mostly interested in and look forward to every release).

[1] https://github.com/tensorflow/tensorflow/releases

We've moved over to Tensorflow from Theano around a year ago. I'm a Software Engineer on the team and here's what I think are advantages from my POV:

1) Transition was fairly straightforward, both APIs' interfaces are more-or-less similar and share some design characteristics.

2) Having said that, TF's API is easier to use and without a doubt a lot easier to read.

3) Consistency: Deploying Theano in different environments surprised me on several occasions with different output compared to the training environment. TF is more consistent on this front (never had such issues).

4) Running multiprocessing with Theano + GPU is a disaster (due to forking) so I end up having to create process pools before initializing Theano. No such issues with TF.

5) TF provides many helpful operators (such as queues and batching ops) as well as monitoring tools (Tensorboard) and debugging tools.

6) Its development is extremely rapid, new releases every couple of months with a lot of improvements and new features every time.

In short, TF is what Theano should have been. A lot of new papers are being developed in TF as well so it helps to understand it.

This comment [1] in Tensorflow's discussion group gives some valid points explaining some of the crazy numbers reported by CNTK.

[1] https://groups.google.com/a/tensorflow.org/d/msg/discuss/Dhy...

That's not really related to the network. When you build a NN, you shouldn't expect different accuracy using different frameworks (given that you're using the same hyper parameters and training/evaluating on the same data).

The point is to be able to work in different windows in the same session, not have two different sessions. I can see how something similar can be achieved in tmux but it would require getting used to a different workflow.

I understand it's possible, but not as easy as screen -x.

We usually have multiple people working in the same screen, each in a different window (on a remote server). Last I checked, this was not straight-forward in tmux (where everyone is switched to the same window whenever someone switches) and it seems to be still the case from what I see in the comparison table. This makes GNU Screen the preferred tool for us, for now.

wow used to use this one before having the app back in the day, good to know it still exists and a perfect alternative for m.facebook.com with messenger support.

That's very helpful, thanks!

I never installed Messenger mainly because I was satisfied with using Facebook app's messaging feature. When it was disabled in the app, I moved to using Facebook in the browser (uninstalled the app). They later removed messaging support in the browser as well so now I request "Desktop mode" when I want to read/write a message. What started as "I don't want to install another app I don't need" became "I will uninstall all apps by Facebook" and now I'm happy I never receive FB-related notifications, I only browse it when I feel like it :)

Same here, can't stand the fact it's taking so long to get Hangouts support on newer versions of Firefox.

I'm on Linux and I don't feel any noticeable speed difference using FF vs Chrome (startup time doesn't bother me at all, my browser is always running and I rarely restart my machine). I've been using FF since the early days and never found enough reason to switch to any other browser and I'm quiet happy with the improvements that have been coming to FF lately and what the future holds.

It's the default browser I'm using right now on Android. Performance is still not on par with Chrome but it did improve since the last time I tried it a few months ago. The ability to install extensions surpasses anything else for me so I'm sticking with FF.

Actually you get a prompt when you try to save a file that has been modified in a different session or when you switch to a similarly modified buffer. But those are pretty reasonable and don't happen that often.

That looks more like emacs behavior though, vim rarely (never?) prompts the user for anything. You enter a command, you get a reaction (either an action like quits vim, or a message indicating something happened) -- which I personally prefer much more than getting annoying prompts.

This was one of the most liked features in Unity for me, and I made sure it existed in GNOME before I made the switch a couple of years ago.

Utilizing multiprocessing for reading and processing jsons (or any type of data) then feeding the output into a shuffle_batch* op works great for me.

Actually this provides a solution for a very serious problem which is usually the main bottleneck in machine learning experimentation (in terms of speed). For me at least, this was a great read and I'm pretty sure I'll be trying this out soon.

tensorflow's py_func allows you to do that, you can also write your own C++ operators, their API interface for that is pretty nice. In fact, you should be able to pretty much write any op you need by combining existing ops (maybe not as efficient as you'd like but yields the results you want).

Thanks for the feedback mamaniscalco, please check back again soon to see if our detection has improved (if you're interested).

As for the the purpose of the tool, one application could be protecting children from online hate/harassment for example.

Hello fwn, interesting examples you have there. We're still working on further improvements to the model at this stage, aiming to detect even more sophisticated examples. Thanks for your feedback and hope you try it again and see improvements soon.

Hello mattbgates, I'm one of the engineers working on the project. Thanks a lot for taking the time to play around with the tool and writing this feedback. We're going to add a feature for giving feedback to moderation (whether the moderated text is actually hate speech or not) to learn from users input.

We're actively working on improving the model so hopefully next time you try it you'll see the improvements!

This should certainly be adopted by official/serious projects (such as albums, movies, events, etc.) with Arabic names written in Latin script. The use of the "chat" form (e.g. "7" for "ح" and "3" for "ع") is just horrid and makes any serious work look childish.

tesserocr releases the Global Interpreter Lock (GIL) while performing some tasks which allows threads to run them in parallel without waiting for the GIL to be available.

It was never tested on Windows, perhaps you can try the Linux steps on Cygwin? I'm interested in seeing how this can work on Windows as I developed this with Linux in mind which was suitable for my use-case