TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Show HN: Deep Learning GUI to Create, Train and Visualize Models in a Browser

216 点作者 0mbre超过 6 年前

17 条评论

WhiteNoiz3超过 6 年前
Cool, project - I really believe that we need more visual tools for building and debugging neural nets.<p>I had a hard time figuring out how to delete a layer (the icons on the left are REALLY tiny and not obvious). Maybe adding a trash can in the graph once you&#x27;ve selected it would be more intuitive (heck maybe add all those icons there too).<p>Also, it would be great if it automatically created &#x27;reshape&#x27; layers for you depending on what you&#x27;re trying to connect or at least suggest them when you try to make a connection that doesn&#x27;t fit. If I&#x27;m going to use a WYSIWG, it should make things easier for you than if you coded it by hand. When I&#x27;m doing AI stuff I find a really big portion of my time is just trying to figure out the right sizes &#x2F; strides &#x2F; padding etc at a given point in the graph. I feel like this would be way easier if it&#x27;s automatically visualized for me or if intelligent suggestions are offered.
hacker_9超过 6 年前
Wow this is great! Makes the process so much more interactive, being able to draw your own digits and it gets recognised is great. Some small issues I ran into for your todo list:<p>1. Load a Model -&gt; Select model -&gt; Dialog had some spelling mistakes.<p>2. Using react tools for chrome, I can see the UI is updating constantly when nothing changes. This is also making the popup flicker when mouse hovering over the graphs. The fix is to have ShouldComponentUpdate return false more!
评论 #19034275 未加载
评论 #19034408 未加载
vinayan3超过 6 年前
Really awesome and will be super helpful as a teaching aide for ML&#x2F;Deep learning classes!
Nashooo超过 6 年前
Hi, the message on mobile has two errors:<p>&quot;Appologies AiFiddle currently does not support mobile device.<p>Do you want to get notified on futur updates ?&quot;<p>&#x27;Apologies&#x27; and &#x27;future&#x27;.<p>Looks promising though, will check back when on desktop.
k4renio超过 6 年前
Very nice concept. This has the potential to be a great teach &#x2F; prototyping tool
ztratar超过 6 年前
I&#x27;m not a machine learning engineer, so I&#x27;ll ask the dumb question: why is visualizing this flow important? What&#x27;s the benefit?
评论 #19033648 未加载
mattlondon超过 6 年前
Really nice - really like it for just fiddling around with models.<p>One missing feature would be an &quot;undo&quot; function - I accidentally deleted a layer and couldn&#x27;t find a way to undelete it?<p>Edit: just found an error when trying to actually train a network:<p>TypeError: Argument 1 of WebGL2RenderingContext.clientWaitSync is not an object.<p>This was doing a very simple MNIST-&gt;dense(200)-&gt;dense(1)-&gt;output model using Firefox 65
评论 #19034278 未加载
评论 #19034428 未加载
amitport超过 6 年前
This is really well done.<p>I&#x27;ve got to mention two missing features that I find important:<p>1 - help popups: it will be nice if terms had a small help icon that can display a little context on hover&#x2F;press (that&#x27;s a quick win, makes unknown things usable, much easier to implement than a full manual&#x2F;interactive tutorial)<p>2 - social login &#x2F; email-only login: you are missing sign-ups (I hate having more passwords).
评论 #19034422 未加载
hnuser123456超过 6 年前
If you want your model to train 20x faster, continuously move your mouse up and down the far left menu, or open the outline and madly expand&#x2F;collapse the options for a layer :D<p>edit: turbo mode: keep v8 at full speed:<p>function sleep(ms) {return new Promise(resolve =&gt; setTimeout(resolve, ms));}; while (true) {console.log(document.getElementsByClassName(&quot;content&quot;).length); await sleep(20)};
评论 #19035989 未加载
edshiro超过 6 年前
Super awesome! I had a similar idea about a year ago but never got to implement it. We need more visual and straightforward ways to create neural network architectures.<p>Can you currently create the architecture and run a given dataset against it? That would be very cool...
评论 #19043409 未加载
abrichr超过 6 年前
The demo video is very compelling! Do you support proprietary datasets?<p>A small suggestion: the audio on the demo video was poor quality and difficult to understand. Perhaps a better microphone or higher bitrate would help.
syntaxing超过 6 年前
This is super awesome! Being able to export this to colab would be amazing. I&#x27;m really looking forward to this feature! Are you willing to share which web framework you used to build your site?
评论 #19035755 未加载
Rainymood超过 6 年前
What is the tech stack used for the website&#x2F;app? Looks great btw.
评论 #19034413 未加载
评论 #19033869 未加载
评论 #19034160 未加载
bicubic超过 6 年前
Awesome work. Now make it emit keras&#x2F;tf&#x2F;torch code!
评论 #19033722 未加载
评论 #19034164 未加载
dadel超过 6 年前
That&#x27;s really great ! I had a look to the available models and played a bit with evaluating canvas and that&#x27;s very nice !
glastra超过 6 年前
Editing a layer and changing the number of filters, stride or kernel size effectively changes the shape of that layer, but it should cascade a re-calculation down the path, which currently doesn&#x27;t happen at least on my screen.
评论 #19036706 未加载
samzer超过 6 年前
This is so slick!