Technology

Image

Automates data normalization by resizing images to 224x224 via Pillow and transcoding audio into uniform 16kHz mono formats.

This workflow automates the heavy lifting of data preparation for multimodal AI. We use Pillow to force images into a 224x224 pixel square (the standard for ResNet and VGG architectures) while maintaining aspect ratio through smart padding. On the audio side, we leverage FFmpeg to transcode diverse formats into 16kHz mono WAV files: this ensures consistent sample rates for downstream spectrogram generation. It is a no-nonsense approach to cleaning noise and unifying inputs before they hit the training loop.

https://pillow.readthedocs.io/

What builders pair with Image

Projects using both technologies. Select a pairing to see a project.

Built with Image

Next 2 shared projects AWS 1 shared project BAML 1 shared project Claude Sonnet 1 shared project Flash 1 shared project Gemini-2 1 shared project

12 more pairings

Gemini 3 1 shared project Gemini Live API 1 shared project Google Gemini Pro 1 shared project Image Processing 1 shared project Images 1 shared project LLM 1 shared project Multimodal Models 1 shared project OpenAI 1 shared project OpenAI API 1 shared project PDF 1 shared project Prompt Engineering 1 shared project RAG 1 shared project

Pairing: Next

Photo from the event — Event photo

Multi-Pass Building Defect Detection: Getting a VLM to Find Facade Defects for Visual Inspections

DC · July 23, 2026

view project | view Next

Recent Talks & Demos

Showing 1-4 of 4

Members-Only

Sign in to see who built these projects

Sign in View FAQ

Multi-Pass Building Defect Detection: Getting a VLM to Find Facade De…

Google Gemini Pro Claude Sonnet

Building NousyBooks - Orchestrating Low-Latency Multimodal Voice Agen…

Building an AI WhatsApp guide for the Valencia Fallas festival

Valencia Mar 17

BAML: Zero-Shot Extraction