The Huging Face Spaces allows you to gain interactive experience with machine learning models, in the article we will consider the best demo work of the Moscow Region.

In this article we will see the best resources using machine training on the Huging Face Spaces.

This Pokémon does Not Exist does not use the Rudall-E model to create Pokemon, and randomized names with Pokemon attributes are generated from the list. To collect rare and unique Pokemon, you need to write your name and press the Sending button. This is a simple web application, but one of my loved ones.

Magma (multimodal expansion of generative models using a thin adjustment based on an adapter) is a model of a visual language for answers to questions about images. Learn more about Magma algorithms in an article by ARXIV.org. To use the ML demo version, you need to load the image and ask a specific question. For example, "Describe the image." Learn more about the use options here.

Animeganv2 is the most popular machine learning application on the Huging Face Space with 515. It gives quick results with an incredible artistic shade. Learn more about the work of generative models Animeganv2 here. To use a demo version, you need to download a portrait, and then choose a style for creating an art in the style of anime.

When I saw the Image Restoration and Colorization demo on Twitter, I thought they must use an example made in Photoshop to demonstrate results. But, when I tried the algorithm myself in a completely new photo, I was amazed at the simplicity and powerful functionality of the application. Demo Gradio asks you to upload a black and white and damaged image, and it will return a color and high-quality photo. You can also play with several options to get different results.

DIT Document Layout Analysis uses a pre -trained Document Image Transformer model to predict the marks in the PDF document. For example, the detection of tables, text, images, etc., a document in PDF format is required for the demonstration, and everything else depends on the selected model for the marking of various parts of the document.

The Chef Transformer demonstration uses the T5-ReCipe-Generation model to create recipes based on the description of cooking, food and ingredients. If you are hungry and you have a limited selection of dishes, enter the ingredients and get a recipe for delicious food. This is my favorite application, because it is visually attractive and really useful)

Arcanegan Video uses the U-Net network, trained in the ARCANE anime set, images are generated using the mixed GAN2 style. Read more about the implementation of the model here. To demonstrate Gradio, you just need to upload a video sample and allow the model to make miracles. The output video will be transformed into the ARCANE anime style.

Rick & Morty Chatbot uses the modified version of Dialogpt, which was trained on the dialogs from the cartoon - Rick and Morty. Just print stupid questions and conduct a conversation with the characters of the cartoon until you get tired.

The OCR for Captcha model was trained using CNN and RNN. To learn more about the training of the model, read the example of the Keras code. The application will ask you to download the image of Kapchi and will return a very accurate alphanumeric text.

FastSpeech2 TTS uses modern architectures of real-time speech synthesis, such as Tacotron-2, Melgan, Multiband-Melgan, FastSpeech, FastSpech2 based on Tensorflow. If you want a natural transformation of the text into speech, try to enter the text and maybe you will be surprised. This application allows you to test various architectures of models.

I like HugingFace Spaces and how the community members come up with unique ideas for web applications. In this article, we examined the list of the ten best demonstrations of machine learning on HF Spaces and learned how these applications work.

#machinelearning #artificialintelligence #ai #datascience #programming #Deechnology #Deeuplearning #bigData #bigdata