Heads fitted on frozen features, one per (task, backbone). Evaluate without training. Each head is valid only against its named backbone.
🤝 Open to Collab
Turhan Can Kargın
turhancan97
AI & ML interests
Computer Vision, Deep Learning, Machine Learning, Robotics, Reinforcement Learning, Control Theory
Recent Activity
updated a collection 3 days ago
VisBench probes updated a collection 3 days ago
VisBench probes updated a collection 3 days ago
VisBench probesOrganizations
Spatial relation probing, given an image and an object triplet, it predicts whether the target is Left / Right / Between relative to reference objects
-
turhancan97/SpaRRTa-probes
Image Classification • Updated • 1 -
SpaRRTA
🧭1Classify object direction (Front/Back/Left/Right) in images
-
turhancan97/SpaRRTa
Viewer • Updated • 149k • 296 • 1 -
SpaRRTa: A Synthetic Benchmark for Evaluating Spatial Intelligence in Visual Foundation Models
Paper • 2601.11729 • Published • 1
Heads fitted on frozen features, one per (task, backbone). Evaluate without training. Each head is valid only against its named backbone.
Spatial relation probing, given an image and an object triplet, it predicts whether the target is Left / Right / Between relative to reference objects
-
turhancan97/SpaRRTa-probes
Image Classification • Updated • 1 -
SpaRRTA
🧭1Classify object direction (Front/Back/Left/Right) in images
-
turhancan97/SpaRRTa
Viewer • Updated • 149k • 296 • 1 -
SpaRRTa: A Synthetic Benchmark for Evaluating Spatial Intelligence in Visual Foundation Models
Paper • 2601.11729 • Published • 1