ann-unsplash-25k / README.md
Ashot Vardanian
Add: Dataset validation and docs
95669d7
|
raw
history blame
873 Bytes
---
license: apache-2.0
---
# 25K Unsplash Images for Search
This is a derivative work based on two existing datasets.
- `images.csv` metadata from [Unsplash](https://github.com/unsplash/datasets), sorted and converted to CSV.
- `images/` in 250x250 resolution by [kaggle/@jettchentt](https://www.kaggle.com/datasets/jettchentt/unsplash-dataset-images-downloaded-250x250).
- `images.fbin` is a binary file with UForm image embeddings.
- `images.usearch` is a binary file with a serialized USearch index.
The original `images.tsv` from Unsplash has been filtered to avoid missing images.
The embeddings and the index can be reconstructed with the `main.py` script.
On the Apple M2 Pro CPU:
- Image vectorization takes 100ms/image, or 10 inferences/second.
- Indexing vectors one-by-one happens at 700 vectors/second speed.
To rebuild the indexes:
```sh
./main.py
```