Loai Houmane的动态

? Project: The app performs on-device face recognition on the live camera feed where users can add images of the persons they wish to recognize. Any other detected face, apart from the database, are detected as 'not recognized' by fixing a threshold on the similarity score. ? Working: When the user selects one or more images from their device, the app detects faces, crops them and creates an embedding with the FaceNet model for each of those cropped faces. These embeddings are then stored in a vector database. Now, on the live camera feed, whenever a face is detected, it is cropped and a FaceNet embedding is generated (we call this the 'query embedding'). This embedding is queried to the vector database, which returns the nearest neighbors. Once we have determined the nearest neighbor, we compute the cosine similarity between the neighbor and the query embedding. If the similarity < threshold, we assume that the face does not belong to any person in the DB, else we assume the face belongs to the person whose face embedding was the nearest neighbor. ? Technologies: Contrary to the earlier projects, I've used the Mediapipe Face Detector for this project which seemed a bit faster than MLKit. TensorFlow Lite is used to run inference on the FaceNet model, with ObjectBox as the on-device vector database. The project follows the clean architecture with modern Android development practices imposed throughout. App download size: 84 MB GitHub: https://lnkd.in/d2Qbw4rt Blog: https://lnkd.in/dbXGsvdp #android #machinelearning #ondeviceml #programming #developers #androiddevelopment #appdevelopment #facerecognition

?? Pro Tip: If you're building a web app and need to share your #MachineLearning model, skip saving it to a file. Instead, create an #API! ?? Why? An API: ? Eliminates environment and security headaches by avoiding dependency and version conflicts. ? Enables access from different languages and platforms—your model is just an HTTP request away! ? Simplifies integration with front-end frameworks, mobile apps, or other services. Ready to get started? Here's a quick guide on how to build an ML API with #FastAPI—get your model online in minutes! ?? #WebDevelopment #APIDevelopment #ML #AI #TechTips #Python #AI

Hello #Connections, ?? Excited to Share My Latest Project! ?? I developed a Text-to-Speech app using Streamlit and gTTS. Users can enter text, which is then converted to speech and played back in the app. The app features a custom background, stylish buttons, and a user-friendly interface. It's a simple yet powerful tool for converting text to audio. I would like to thanks to my guide, KODI PRAKASH SENAPATI Sir for his invaluable support and guidance. Check out the project on GitHub for more details : https://lnkd.in/gEaCYWtU #TextToSpeech #Streamlit #gTTS #DataScience #AI #Python

??Visual Question Answering App - test with images?? This is the second post related to this app. I just changed the UI a little bit, and this video is all about testing the deployed app. Thanks to Gradio for their exceptional SDK and Hugging Face for easy deployment. ?? Link to app: https://lnkd.in/dM5BM4XB ?? GitHub: https://lnkd.in/dRjKuCM3 #AI #Python #Machinelearning #Finance #Medicalimage #Imageprocessing #Computervision #Visualquestionanswering #Gradio #Hugginface #Datascience #Chartanalysis #Businessanalysis

Elevating My Data Analysis: From Local to Global ???? Transforming my local Python and Excel-based #USEconomicData analysis into a sleek, user-friendly web app! ??? Thanks to the power of #GenerativeAI, I'm revolutionizing my workflow. Now, I can access and interpret critical economic insights with ease from anywhere. The best part? I built this powerful tool through #NoCodeProgramming, making data-driven decisions more accessible than ever. #DataAnalysis #WebApp #TechInnovation #EconomicInsights #ProductivityBoost Who else is leveraging AI to streamline their data processes? Share your experiences below! ?? PM me if you like the web app.

?? Introducing My New Project: AI-Powered Web Scraper (v1) ?? I’m excited to share the first version of a tool that combines web scraping and AI to extract specific data from websites! Here's how it works: 1?? Enter a website link 2?? Specify the data you need 3?? Get results instantly, powered by LLaMA 3.2-11B ?? Built with Python, Streamlit, and llama 3.2 , this app simplifies data extraction for everyone. ?? Try it here: https://lnkd.in/dCyDRZvk #AI #WebScraping #LLM #Python

? Sentence embeddings now on Android! The app uses the popular all-MiniLM-L6-V2 from Hugging Face sentence-transformers to produce 384-dimensional embeddings for given sentences. The inference time, in the video below, on a 32-bit low-end Android device is about 700-800 ms for both sentences, without any optimizations (like NNAPI or other delegates). The downloadable size of the app is 112 MB with ~90 MB taken by the model (see 1st comment). In my previous project, Android-Document-QA, I used Mediapipe's Text Embedder which utilizes Google's Universal Sentence Encoder model. But its performance on determining semantic similarity was not upto the mark, though I do not have any benchmarks to prove the same. This led me to bring sentence-transformers on Android, that had excellent performance in a similar Python project of mine. A highly performant sentence embedding model is useful for on-device RAG applications and text-similarity analysis. The blog and GitHub repo will be shared soon, describing the process of bringing sentence-transformers on Android! #machinelearning #android #androidappdevelopment #ondeviceml #development #developers #rag #softwaredevelopment

Did you miss my final #RAGHack stream, "Evaluating your RAG chat app"? Catch the recording here: ?? https://lnkd.in/gSd32wXm I covered: ?? What does it mean for a RAG answer to be high quality? ?? Which factors affect RAG answer quality, and which affect it the most? ?? Using promptflow-evals SDK to run GPT evaluators in Python ?? Using my ai-rag-chat-evaluator CLI tools to generate synthetic ground truth data, run batch evaluations, and compare metrics across runs ????♀? Strategies to make sure your app can say "I don't know" at the appropriate times Grab slides from https://lnkd.in/g-9xSQvV #rag #openai #python #evals

AI Artifacts App: An Open Source Version of Anthropic Artifacts that can Analyze Python Code, Generate HTML/CSS/JS and Next.js Code Many developers face the challenge of safely executing AI-generated code. Running such code locally can pose security risks and may require extensive setup. Additionally, there’s a need for a tool that can support multiple programming languages and ... https://lnkd.in/ebRYn6md #AI #ML #Automation

?? Introducing the Cartoonify App! ??? I'm excited to share my latest project: the Cartoonify App, where advanced image processing meets user customization! With this app, you can transform regular photos into eye-catching cartoon-style images, add creative enhancements, and adjust effects to create your ideal look. ?? Project Highlights: 2D/3D Cartoon Effect for added depth and vibrancy Enhanced Vignette and Background Blur for professional touches Customizable Filters for brightness, contrast, sharpness, and more User-Friendly Registration & Login System to secure user data This project was built using a combination of Python, OpenCV, Streamlit, and SQLite for backend and image processing. It has been a fun, challenging, and rewarding experience to integrate these tools and develop an intuitive UI. Check it out, and feel free to connect with me for collaboration or feedback! ?? #ImageProcessing #Python #ComputerVision #Cartoonify #OpenCV #Streamlit #TechProjects #AIArt #DigitalTransformation #DataScience #MachineLearning #SQLite #AppDevelopment #Innovation #UserExperience #PhotoEditing #CreativeAI #AIProjects 4o