Made siglip-256 as default classifier

Added siglip multiple res models

Removed object detection tab

Added dfine all models and removed jina and siglip2

Added threshold slider for siglip models

Updated siglip labels

Fixed siglip2 issue

Fixed ref bug

Added siglip and siglip2 for classification

Merge branch 'main' of https://huggingface.co/spaces/Napron/small_object_detection

Modified the pipeline to run DFINE on small crops again

Expanding by 10% person/car crop to make sure objects centers are still within person/car bbox

Modified default thresholds for display and gap

Added gap value to the interface

Added person crop threshold 0.9

Added DFine large model as option and removed nomic

Expanding smaller objects inside person/car crop by 60%

Add Threshold slider for min display confidence (default 0.7)

D-FINE: show only groups with known objects, display conf >= 0.7

D-FINE: person/car crop galleries, known-object bboxes only

Update dfine_jina_pipeline.py

Update app.py

Update dfine_jina_pipeline.py

Update app.py

modified nomic encode images

Update nomic_fewshot.py

Update nomic_fewshot.py

Update jina_fewshot.py

Update jina_fewshot.py

Update jina_fewshot.py

Update jina_fewshot.py

Update requirements-lock.txt

Update jina_fewshot.py

Update jina_fewshot.py

Update jina_fewshot.py

Update jina_fewshot.py

Update jina_fewshot.py

Update Dockerfile

Update jina_fewshot.py

Update requirements-lock.txt

Update requirements.txt

Update requirements-lock.txt

Log device, Jina CPU warning, pin revision