---

Introduction

• This project implements scalable Ray-based enrichment of datasets with CLIP embeddings to enable retrieval with image and text queries, cluster analysis,and deduplication.

• The notebook provides code to read in images, generate embeddings, and publish the dataset via Hugging Face.

---

Sample search results

---

Dataset

Source

All images were sourced from Francesco/insects-mytwu.

Link

https://huggingface.co/datasets/hkanade/insect_image_retrieval/

Sample usage

from datasets import load_dataset

ds_new = load_dataset("hkanade/insect_image_retrieval")
ds_new["train"][0]["image"]