CoreSole™ Shoeprint Retrieval Model
State-of-the-art Vision Transformer AI model for detecting and accurately matching footwear impressions in crime scene photos.
Model Details
Model Description
- Developed by: Coreviz
- Shared by: Coreviz
- Model type: Vision Embedding Model (based on ViT)
- License: Consult with us.
- Finetuned from model openai/clip-vit-large-patch14
Model Sources [optional]
- Homepage CoreSole
- Paper Coming soon.
- Interactive Space HuggingFace Space
- Demo CoreSole Live Demo
Uses
For use in assisting forensic professionals in sorting through crime scene footwear impression imagery.
Direct Use
Returns accurate embeddings for shoeprints based on tread patterns, ignoring the impression's surface (will have closer similarity for shoeprints that have similar patterns regardless of whether the shoeprint is on dust, dirt, blood or a scanned shoe sole).
Downstream Use [optional]
Bulk tagging and similarity search is available through the Rekon forensic platform.
Out-of-Scope Use
Not for use in court proceeding. Not admissible in court. Does not replace expert opinion and does not perform analysis or exact shoeprint matching. For use only as a lead generation tool after thoroughly verifying potential matches. False negatives and false positives are possible and expected.
How to Get Started with the Model
Talk to us for access to the Rekon forensics platforms. Currently running beta pilot programs.
Model tree for coreviz/coresole
Base model
openai/clip-vit-base-patch32