22 days ago
-
Add embed(), config-selected backbone width, depth decoder and detection preprocessing
-
Report depth on the NYU Eigen test split
-
Return maps at the input shape, batch NMS, ship ADE20K names, add compile() and evaluate.py
-
Fix non-RGB inputs, offline class names, depth thread safety, perceive timings, the export check and the INT8 config
-
Fix depth at odd patch grids and the detection ONNX export grid; shorten the model card
-
argus.py: factor padding_mode through DPT blocks; add depth crop_border kwarg; batched correspond. README: align with shipped 3M cofiber detection head, drop FCOS framing, fix file sizes/param counts, add real IN1k val comparison, document qkv-bias choice. eval JSON: strip personal paths.
-
README: remove session-narrative paragraph
-
README: document reflect-pad DPT swap (0.646 -> 0.314 RMSE on 5050-frame flat val)
-
argus.py: reflect-padded DPT convs, drop post-hoc border crop
-
add imagenet_classes.json
-
config: extract class_ids/class_names to imagenet_classes.json
-
argus.py: correspond() returns dense {matches, scores, grid} dict; lazy-load class names
-
README: bulletize architecture, drop duplicate DPT paragraph, refresh classification baselines
-
Add per-model latency to RF100-VL benchmark table
-
Tighten RF100-VL cross-domain benchmark section
-
Add RF100-VL cross-domain detection benchmark
-
update repository
