Skip to content

akritisshetty/trendlens

Repository files navigation

trendlens

Build Instructions (TrendLens Pipeline)

Run the following scripts in order:

python generate_metadata.py        # Generate metadata for each post
python generate_embeddings.py      # Generate CLIP embeddings for images + metadata
python generate_umap.py            # Perform dimensionality reduction (UMAP)
python convert_npy_to_csv.py       # Convert .npy files to .csv for readability

Output

All outputs are stored in trendlens_outputs/:

trendlens_outputs/
├── embeddings.csv
├── embeddings.npy
├── embeddings_checkpoint.npy
├── metadata.csv
├── metadata_checkpoint.csv
├── smpd_metadata.json
├── umap_2d.npy
├── umap_10d.npy
├── umap_2d.csv
├── umap_10d.csv
└── umap_scatter.png

About

A system that detects emerging visual trends on social media by analyzing images directly, not just text.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages