Go back

Coding Projects by Bhargav

AI-Powered Image-to-Text Conversion Project – Part 2

Students will apply AI-powered tools to generate captions, descriptions, and summaries from images. By expanding on a simple image caption, they will utilize a text generation model (GPT-2) and a text-to-image model (Stable Diffusion) to create a complete, interactive workflow for text-to-image conversion. They will practice generating descriptive text, summarizing the content, and exploring creative applications.

Snap-to-Caption

Practice generating a basic image caption and optionally expand it using GPT-2, with simplified code and minimal user interaction.

Multi-Image AI Captioning Challenge

Extend the basic image captioning script to handle multiple images, track captions in a report, and prompt for user input. Strengthen Python fundamentals while exploring deeper AI-driven captioning workflows.

Vintage Photo Healer

Strengthen students’ skills in AI-driven inpainting by restoring damaged areas of a single vintage photograph with Stable Diffusion, using masks and descriptive prompts exactly as practiced in class.

AI Image Stylist: Your Signature Look

To reinforce students’ understanding of AI image generation and post-processing using Stable Diffusion and Pillow by creating a themed image enhancement pipeline.

Text-to-Image Payload Customization

Extend the text-to-image script to fine-tune prompt parameters. Modify the payload structure for stylistic nuances, adjust response settings, and explore optional fields to gain deeper control over generated images.

Next Word Notebook

Build the next-word guesser of a tiny keyboard: count which word follows which in a short text, turn the counts into probabilities, predict the most likely next word (or fall back to the most common word), then count the mistakes. Time: about 20 minutes.

Note Sorter

Build the sorting box of a school notice board: count the words of each box, compare a new note with every box using cosine, pick the best box or say "unknown", then count the mistakes your sorter makes. Time: about 20 minutes.

Recipe Box Search

Build the search box of a recipe app: index every word of five recipes, score each recipe for a search, pick the best one, then count the mistakes your search makes. Time: about 20 minutes.

Pet Shop Search

Build the "Did you mean?" box of a pet shop: count the fixes between a search and each pet, suggest the closest pet, then count the mistakes your search makes. Time: about 20 minutes.

Canteen Comment Box

Read canteen comments and mark each one HAPPY, UNHAPPY or OK with word points and a "not" flip, then count the mistakes your program makes. Time: about 20 minutes.

Bakery Help Desk Sorter

Sort a bakery's help desk messages into URGENT and NORMAL with a weighted word score, then count the mistakes your sorter makes. Time: about 20 minutes.