Nitin Tiwari

Llama3-on-Mobile. This repository is an implementation of quantizing and converting the Llama3-8B-Instruct model weights and deploying it on Android for on-device inference.

Sketch2Vid. This repository is an implementation of converting sketches into lively videos using Google's Veo 3 model.

Gemma-on-Android. This project is an implementation of fine-tuning the Gemma 2b-it model on a custom dataset and deploy the model on Android.

YOLOv10-LiteRT-Android. This repository is an implementation of converting the YOLOv10 object detection model to LiteRT (.tflite) format and deploy it on Android using Google AI Edge for on-device inference.

TensorFlow.js-Custom-Object-Detection. An E2E custom object detection browser-based application using TensorFlow.js.

Fine-tune-IDEFICS-Vision-Language-Model. This repository demonstrates the data preparation and fine-tuning the IDEFICS Vision Language Model.

Virtual-Try-On-Imagen3. This repository is an implementation of virtually trying on different outfits by image editing and inpainting using Imagen 3 model.

Medicine-Scan-with-Gemini. PharmaScan is an Android application that leverages Gemini Pro Vision model to identify medicines and provide their details such as usage, dosage, diagnosis, etc. on-the-go.

YOLOv3-Custom-Object-Detection. An E2E tutorial on custom object detection using YOLOv3 with Transfer Learning on Google Colab.

PaliGemma. This repository contains examples of using PaliGemma for tasks such as object detection, segmentation, image captioning, etc.

PaliGemma-Android-HF. This repository is an implementation of inferring the PaliGemma Vision Language Model on Android using Hugging Face-Gradio Client API for tasks such as zero-shot object detection, image captioning and visual question-answering.

Cartoon-Classification-on-Android-using-TF-Lite. A fun image classification Android app that classifies cartoons using TensorFlow Lite.

TFJS-TFLite-Object-Detection. This repository is an implementation of object detection to detect waste in real-time directly on the browser using the TFJS-TFLite Web API.

FarmScan-using-Gemini. FarmScan: Farmer's Assistance is an implementation of using the Google Pro Vision model API on Android to recognize the freshness of fruits/vegetables, its approximate market value, shelf life, and a lot more insights to help farms plan cultivation/selling of crops better.

Custom-Object-Detection-on-Android-using-TF-Lite. An end-to-end tutorial to train a custom object detection model and deploy it on Android using TensorFlow Lite.

Video-Game-Control-using-Pose-Classification-and-TensorFlow-Lite. An E2E tutorial to create a video game controller by creating a custom pose classifier using TF Lite.

Stable-DiffusionXL-using-DreamBooth-and-LoRA. This project is an implementation of fine-tuning an SDXL model using DreamBooth and LoRA on custom data of interior rooms to generate designs for your home.

Custom-Text-Classification-on-Android-using-TF-Lite. An E2E custom text classification application for Android using TensorFlow Lite.

YOLOv3-to-TensorFlow-Lite-Conversion. Convert your pre-trained YOLOv3 models into its corresponding TensorFlow Lite version and test the resulting TF Lite model.

Object-Detection-using-Gemini. This repository demonstrates how to do object detection on any object using Gemini API.

PaliGemma2-ONNX-Transformers.js. This repository is a step-by-step implementation of converting and quantizing the PaliGemma 2 Vision Language Model to ONNX weights, and inferencing it on the browser using Hugging Face Transformers.js.

NSTiwari.

Inference-Gemma-3. This repository is an implementation of inferencing the Gemma 3 model for multiple tasks using Hugging Face Transformers.

GTA-Vice-City-Object-Detection-using-TensorFlow. A cool project where video game meets Computer Vision. An end-to-end tutorial to train and deploy an object detection model on Grand Theft Auto: Vice City.

DCGANs-using-Keras-and-TensorFlow. This repository demonstrates the implementation of DCGANs to generate fake human images trained on the CelebA dataset using Keras and TensorFlow.

Sketch2Paint-Gemini. This repository is an implementation of converting hand-drawn sketches to creative paintings using Gemini Flash 2.0 Image Generation model.

Hindi-Character-Recognition-on-Android-using-TensorFlow-Lite. Jupyter Notebook

GoogleMapsStreet2Anime. This project is an implementation of transforming places in Google Maps Street View to anime using Gemini.

Build-with-AI-Vizag-2024. This repository contains resources for my talk on Gemini & Gemma during Build with AI Vizag 2024.

Mobile-VisionLM. This repository is an implementation of converting, quantizing and deploying the a vision language model on mobile for on-device inference.

ML-Session. Resources for ML session.

Vision-API-on-Google-Cloud-Platform. This repository provides practical demonstrations of Google Cloud Vision API usage for tasks like face detection, object localization, landmark recognition, text detection, etc.

Virtual-Try-On-Imagen4. This repository is an implementation of creating traditional attires based on user-selected locations, leveraging Imagen 4 and Virtual Try-On APIs for a culturally immersive experience.

CommunityTalks. Talks given by me at various seminars, workshops and meet-ups.

3D-Objects-with-Imagen3-SV3D. This repository implements the creation of 3D visualizations of objects using Stable Video 3D (SV3D) from creative images generated with Imagen 3.

Gemini-Build-with-AI-Asaba. Resources for Gemini & Gemma talk - GDG Asaba.

Multifunctional-Calculator. C

cookbook. Examples and guides for using the Gemini API

Img2Anime-Stable-Diffusion. This repository is an implementation of creating photo journal of anime pictures from your image gallery using Stable Diffusion XL Turbo and captioning using Gemini.

Segmentation-using-Gemini. This repository is an implementation of referring expression segmentation on images and videos using Gemini 2.5 model.

WhatsApp-Bot. WhatsApp Bot is a Python-based chatbot for WhatsApp to automate the process of sending messages to one or more person or a group.

devlibrary. Source code for Google Dev Library

Human-Pose-Estimation-using-OpenPose-and-OpenCV. Detect human poses in real-time using OpenPose and write the detected video or image using OpenCV.

2D-Colour-Barcode-Steganography-using-MATLAB. 2D Colour Barcode Steganography is used to hide the data in the form of coloured barcode which is then retrieved back by processing the image of the coloured barcode.

Visual-Data-Extractor-using-TensorFlow. A Computer Vision application built using TensorFlow for businesses that require data pipelines for extraction of visual data (graphs and charts) from images and dashboards.

Android-MySQL-connectivity. Java