Show HN: Open-Source Zero-Shot Image Model Server Enabling Model Feedback https://ift.tt/y7MPRoQ

Show HN: Open-Source Zero-Shot Image Model Server Enabling Model Feedback Hi everyone! Here is an open source implementation of a decently performant server hosting zero-shot image models (CLIP for image classification, OWL-ViT-ST for object detection), with an extra algorithm to allow users to give the models feedback when they make mistakes! We built a company off this flavor of tech two years ago and have clients who are currently using our commercial API. We are now moving on to other projects but want to make sure our clients still have access to the approaches that they've grown to rely on, so we're open sourcing a simple implementation that they'll be able to use after we've shut down our hosted API! I used to work at a robotics startup. After a while it seemed clear that the biggest limiting factor in our ability to ship new models wasn't innovation on model architecture, it was access to relevant, high-quality training data. Around that time CLIP was released, which got me thinking about the idea of having models with world-knowledge baked in so as to reduce the amount of training data required. A year later when Stable Diffusion dropped, my cofounder Ben Brooks and I took the plunge and founded DirectAI, where we worked on building ways to get performant models without collecting any training data, using the knowledge stored in pretrained models instead. In this implementation, we replace the linear classification head typically used in zero-shot image classifiers with a modified nearest neighbors method that lets you use multiple examples (both positive and negative) per-class to make sure the decision boundary the model is using is more aligned with what you had in mind. Our clients have found it very useful for things from interior design to content moderation to sports analytics, building models that are either too niche to be supported by a traditional cloud-hosted computer vision API or are subtly different from the models that existing cloud APIs host. For example, one of our clients wants to filter out all images containing alcohol. Hive has an API for that, but Hive explicitly allows red solo cups that don't obviously have anything alcoholic in them, whereas our client wanted to filter those out too! Feedback is welcome! There are still bugs in the Gradio frontend / codebase in general, but I have a deadline and need to be working on new stuff at a new job starting Monday so I thought I would just go ahead and get it out there! I've never tried to publish a real open source piece of code before and I must admit I am quite nervous! https://ift.tt/a7qXJYS October 20, 2024 at 12:21AM

Show HN: I made a site to quick identify any plant and learn how to care for it https://ift.tt/z27qGAl

Show HN: I made a site to quick identify any plant and learn how to care for it I'm still working on it, curious but what feature may be most valuable for you? What do you think about personalised care instructions, and an interactive chat feature for each plant. https://frondly.app/ October 20, 2024 at 02:30AM

Show HN: Contagious Beliefs–Simulating Political Alignment https://ift.tt/nAryOIl

Show HN: Contagious Beliefs–Simulating Political Alignment This is a simulation that allows you to test how beliefs are adopted based on pre-existing beliefs and how they spread through a population. It is entirely editable, so you can test your own parameters based on your own data or intuitions. A detail methodology can be found here: https://ift.tt/eqmY3t0... https://ift.tt/VtvSew0 October 20, 2024 at 12:23AM

Show HN: No Setup Needed Play NES, SNES, Sega, Atari Games Directly in Browser https://ift.tt/OjEgMe2

Show HN: No Setup Needed Play NES, SNES, Sega, Atari Games Directly in Browser Tired of the hassle of setting up classic NES, SNES, Sega, Arcade, and Atari games on modern systems? I have built something way simpler - browser based emulation. Key Features: Instant Play: No downloads or complex configurations required. Extensive Library: Explore iconic titles like Super Mario Bros., The Legend of Zelda, Sonic the Hedgehog, Pac-Man, and Street Fighter II. Seamless Emulation: Enjoy smooth gameplay powered by advanced browser technology. Cloud Saves: Never lose your progress again with cloud saving. Let me know your thoughts/suggestions and help bring back the glory days of old age gaming! https://ift.tt/a2z4nVQ October 19, 2024 at 05:25AM

Show HN: I built a tool that helps people contact you without spam https://ift.tt/K48rAez

Show HN: I built a tool that helps people contact you without spam https://hi.new October 19, 2024 at 05:16AM

Show HN: Run, learn, and debug x86-64 Assembly code directly from your browser https://ift.tt/xNaYR5E

Show HN: Run, learn, and debug x86-64 Assembly code directly from your browser I recently created this project to teach assembly, and to have an assembly playground in the style of codepen/jsfiddle to embed in my blog posts. It's based on a wasm version of the blink x86-64 emulator, so everything runs client side with no backend server and with a runtime of only ~200Kb https://x64.halb.it/ October 18, 2024 at 11:02PM

Show HN: Did – Decentralized global social feed https://ift.tt/HLgvjYh

Show HN: Did – Decentralized global social feed https://ift.tt/b8CYjZe October 18, 2024 at 02:27AM

Show HN: Thread Archive – free tool to save Slack community data https://ift.tt/HujZ8wE

Show HN: Thread Archive – free tool to save Slack community data Slack's new content deletion policy is a nightmare for 1000s of communities on Slack. It means that all messages older than one year will be permanently deleted from their servers. Thread Archive gives you a simple way to save all your Slack community's chats in an accessible archive. https://ift.tt/7mKjAzX October 18, 2024 at 03:30AM

Show HN: React-guitar-chord, a react component for chord name to guitar diagram https://ift.tt/iKaS2Is

Show HN: React-guitar-chord, a react component for chord name to guitar diagram I made a pretty simple react component that takes in a chord name and spits out a guitar chord diagram. So far it has no regard for playability or bar chords or any alternate positions, and doesn't ensure that all the notes of a complex chord are played. It has a bias towards open notes so with enough alterations it tends toward an open strum. Still, it's pretty accurate for simpler 3 or 4 note chords. I think my next step will be ensuring that 4 or less fretted notes are included in a chord diagram, using bar chords and moving up the neck if needed. https://ift.tt/wviWSV8 October 18, 2024 at 12:29AM

Show HN: I made a webapp where you can share your projects https://ift.tt/qetisvE

Show HN: I made a webapp where you can share your projects Hi guys I made this simple web application, you can use it to share your projects. Your project will appear only if you visit the projects of the other users. In the 'you' section, add an app ( name, link, [icon_link]) and apply the starting points you got, to the app so that all the other N users will see it (where N indicates how many points you applied to the app). IT'S TOTALLY FREE NO PASSWORD NEEDED FOR LOGIN https://appearnow.me October 17, 2024 at 10:18PM

Show HN: Citizen – A Node.js web app framework for fans of classic app servers https://ift.tt/7lMR41A

Show HN: Citizen – A Node.js web app framework for fans of classic app servers Citizen is a web application framework I started building years ago when I was first learning Node. I've added features and improved stability/performance continuously and finally decided it was worthy of 1.0 status. I think it might appeal to devs like me (old guys) who came up on app servers like ColdFusion, .NET, PHP, etc. and prefer the server do the heavy lifting. It doesn't dictate anything about your front end, and instead tries to be as flexible as possible in providing whatever output the client might need, whether it's a single fully-rendered HTML view, partials, JSON, or plain text. Only 2 direct dependencies for terminal commands (commander) and file watching (chokidar). Built-in functionality includes zero-configuration server-side routing, session management, key/value store (cache rendered views, controller actions, objects, and static files), simple directives for managing cookies, sessions, redirects, and caches, and more. It's been in continuous use on at least one site that receives hundreds of thousands of page views per month, running months at a time without the process crashing. Fairly stable. Appreciate any comments/feedback/contributions. https://ift.tt/Vx1hPr4 October 17, 2024 at 12:18AM

Show HN: Semantic Search for Idea Validation https://ift.tt/ni3GXgt

Show HN: Semantic Search for Idea Validation https://brainwave.vc/ October 16, 2024 at 11:38PM

Show HN: Podcastfy AI – Open-source tool to generate AI audio conversations https://ift.tt/gzqDv4b

Show HN: Podcastfy AI – Open-source tool to generate AI audio conversations https://ift.tt/zMUlXKE October 16, 2024 at 01:18AM

Show HN: 4B+ DNS Records Dataset https://ift.tt/Tg71MtG

Show HN: 4B+ DNS Records Dataset Hi HN, I've been working on building a pipeline to create a DNS records database lately. The goal is to enable research as well as competitive landscape analysis on the internet. The dataset for now spans around 4 billion records and covers all the common DNS record types: A AAAA ANAME CAA CNAME HINFO HTTPS MX NAPTR NS PTR SOA SRV SSHFP SVCB TLSA TXT Each line in the CSV file represents a single DNS record in the following format: https://ift.tt/rvDGJdT Let me know if you have any questions or feedback! https://ift.tt/EOoV3iU October 16, 2024 at 12:56AM

Show HN: Turn any YouTube video into a summary, quiz, & flashcards https://ift.tt/hXYD2to

Show HN: Turn any YouTube video into a summary, quiz, & flashcards YouTube is the greatest learning resource in the world. Yet it doesn't provide any tools to help you learn from videos. I built a simple app that turns YouTube videos into structured learning material using AI. It generates a summary, quiz and flashcards from video content, meaning you can learn from anything (well, at least any YouTube video). Try it out. Would be happy to keep building features based on feedback! https://asterlab.io/ October 14, 2024 at 11:15PM

Show HN: Those Who Die as Cattle https://ift.tt/YGPLxTe

Show HN: Those Who Die as Cattle What if a game gave you the backstory of everyone you killed or got killed? I feel like I've heard this question asked before, but I've never seen it done. For Ludum Dare 56 (a game jam), I spent 48 hours building a prototype called Those Who Die As Cattle, a WW1 trench warfare simulation game. I used ChatGPT 4o-mini to generate details for the 12,000 soldiers involved in the game. Notably, I had to supply names because it couldn't introduce sufficient variety. I also had to add randomly generated adjectives for the soldiers; otherwise, the prompts were too similar, even with different seeds and temperatures. Initially, I used Claude (which I usually prefer), but it made every soldier a former carpenter’s apprentice. Since this is a game and your actions should matter, instead of generating one block of text, I had it generate JSON by providing a template with cues for different things that might happen to the soldier. I used these to dynamically assemble a story for each casualty. For example: { "briefBio": "Tell me a little bit about the solider", // Putting this first is critical to introduce variety … "lastThoughtsArtilleryDead": "What are his last thoughts as he dies from an artillery strike" , … "effectOnOthersShotInjured": "What are the long term impacts on the people who care about him when he is seriously injured by a bullet wound", } The prompt template can be found here: https://ift.tt/WX8Ecrd... The underlying simulation was programmed using Unity ECS (Entity Component System) and simulated, to some degree, each soldier and artillery piece. Casualties were taken in entity order, so there's some predictability regarding when each name gets added to the list, though it isn't completely deterministic. The area where I skimped the most was the UI, but the game is in a playable state if you're willing to look at the center column and figure out what the numbers mean. A final notable learning is that Unity UI toolkit (their modern, web dev like solution) performs significantly better on the web than in the native linux build, at least on the 2022 LTS version of Unity. This suggests that they may be leveraging the rendering efficiencies built into browsers. https://ift.tt/TaWZJt0 October 14, 2024 at 09:32PM

Show HN: I built Bazaari.io to make launching online stores simple and fast https://ift.tt/r5vK1zT

Show HN: I built Bazaari.io to make launching online stores simple and fast https://bazaari.io/ October 14, 2024 at 03:29AM

Show HN: Book to Classical Music Soundtrack https://ift.tt/UZc2J6f

Show HN: Book to Classical Music Soundtrack Hey, HN, literature and classical music are two of my biggest passions, and I always found it interesting how some classical pieces fit to the stories they are barely connected with. E.g. some of Tchaikovsky's music will always evoke the world of Harry Potter for me; music of Sibelius takes me to the same landscapes as writings of Tolkien - of course, these clearly result from John Williams' and Howard Shore's direct inspirations from the romantic composers when creating the soundtracks for the movie adaptations. For a more niche example, music of Olivier Messiaen fits in my view the book L'Écume Des Jours by Boris Vian - grandiose movements of the Turangalila symphony for the first half of the story, the expressionism and sadness of Vingt Regards the latter. The examples are numerous. My recent idea was to somehow automate this pairing of literature and music. It works quite simply - at first, I collected a large dataset of classical music pieces with text descriptions, which I transformed into embeddings. When generating a soundtrack, I use an LLM to create a description of what kind of music would be suitable for the story, and the embedding of this description is used to fetch the most relevant tracks. While the resulting playlist can barely be understood as "soundtrack" at this point, it can be a pleasant background playlist for reading the book. Thanks in advance for any suggestions or feedback :) https://ift.tt/B3PCMft October 14, 2024 at 02:58AM

Show HN: I made a git rebase TUI editor https://ift.tt/AU7FBgN

Show HN: I made a git rebase TUI editor I use interactive rebase quite often, and particularly like the editor bundled with IntelliJ. But I do not always work with IntelliJ, and am not 'fluent' with Vim, so I tried to replicate roughly the same rebase experience within a TUI. I used a small TUI OCaml project i made last year. The notable features are: - Move commits up and down, fixup, drop - Rename commits from the editor (without having to stop for a reword during the rebase run) - Visualize modified files along commits - 'Explode' a commit ,creating a commit for each modified file (a thing I found myself doing quite often) Feedbacks (both on the tool and the code) and contributions welcome, hope it could fit other people needs too ! https://ift.tt/wDL7M3J October 14, 2024 at 01:33AM

Show HN: The fastest Ethereum graph explorer (OpenSource) https://ift.tt/Xnusvq0

Show HN: The fastest Ethereum graph explorer (OpenSource) https://dictynna.com/ October 14, 2024 at 12:51AM