Pico-Banana-400K: A Large-Scale Dataset for Text-Guided Image Editing (Oct 2025)

Title: Pico-Banana-400K: A Large-Scale Dataset for Text-Guided Image Editing (Oct 2025)
Link: http://arxiv.org/abs/2510.19808v1
Date: October 2025

Summary:
Pico-Banana-400K is a large-scale, high-quality dataset of approximately 400K real-image-based text-guided image edits. It addresses the lack of robust datasets for instruction-based image editing by leveraging Nano-Banana for generation and Gemini-2.5-Pro for quality filtering, ensuring diverse edit types via a 35-type taxonomy. The dataset includes single-turn edits, 72K multi-turn editing sequences for complex scenarios, 56K preference pairs for alignment, and dual instruction formats (long/short). It serves as a comprehensive resource for training and benchmarking next-generation image editing models.

Key Topics:
Text-Guided Image Editing
Large-Scale Dataset
Multimodal Language Models
Dataset Construction
Quality Control
Multi-Turn Editing
Preference Learning
Instruction-Based Editing
OpenImages

Chapters:
00:00 - Data Bottleneck in AI Editing
01:34 - Pico Banana Dataset Features
02:56 - Dataset Creation & Model Limits
05:12 - Automated Quality Judgement
07:02 - Learning from Failures & Dual Prompts
08:59 - Advanced Research Subsets
11:14 - Model Performance & Challenges
12:59 - Future of Geometric Editing

Stock video credits:
Colin Jones - https://www.pexels.com/@larchmedia
StefWithAnF - https://www.pexels.com/@stefwithanf-1...
Anthony 🙂 - https://www.pexels.com/@inspiredimages
Pachon in Motion - https://www.pexels.com/@pachon-in-mot...
KATRIN BOLOVTSOVA - https://www.pexels.com/@ekaterina-bol...
Yaroslav Shuraev - https://www.pexels.com/@yaroslav-shuraev
Pixabay - https://www.pexels.com/@pixabay
Engin Akyurt - https://www.pexels.com/@enginakyurt
Soumya - https://www.pexels.com/@soumya-1446957
Mikhail Nilov - https://www.pexels.com/@mikhail-nilov
Stas Knop - https://www.pexels.com/@stasknop
Kindel Media - https://www.pexels.com/@kindelmedia
Silviu Din - https://www.pexels.com/@silviu-din-16...
Pressmaster - https://www.pexels.com/@pressmaster
Danil Shostak - https://www.pexels.com/@danil-shostak...
Trippy Lagoon - https://www.pexels.com/@trippy-lagoon...
Charlie Mounsey - https://www.pexels.com/@charlie-mouns...
cottonbro studio - https://www.pexels.com/@cottonbro
Dan Cristian Pădureț - https://www.pexels.com/@paduret
Oleg Gamulinskii - https://www.pexels.com/@oleg-gamulins...
crazy motions - https://www.pexels.com/@crazy-motions...
Kelly - https://www.pexels.com/@kelly
Pavel Danilyuk - https://www.pexels.com/@pavel-danilyuk
José Alfredo Munguía Lira - https://www.pexels.com/@rectorretro
@svetjekolem - https://www.pexels.com/@svetjekolem